thevar1able/ClickHouse
1
0
Fork 0
mirror of https://github.com/ClickHouse/ClickHouse.git synced 2024-11-24 08:32:02 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'upstream/master' into HEAD

Browse Source
This commit is contained in:
Anton Popov 2022-09-08 01:52:52 +00:00
commit 7c12b448b8
267 changed files with 5591 additions and 2250 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

228
.github/workflows/pull_request.yml vendored
View File

@ -1254,6 +1254,228 @@ jobs:
# shellcheck disable=SC2046
docker rm -f $(docker ps -a -q) ||:
sudo rm -fr "$TEMP_PATH"
FunctionalStatelessTestS3Debug0:
needs: [BuilderDebDebug]
runs-on: [self-hosted, func-tester]
steps:
- name: Set envs
run: |
cat >> "$GITHUB_ENV" << 'EOF'
TEMP_PATH=${{runner.temp}}/stateless_s3_storage_debug
REPORTS_PATH=${{runner.temp}}/reports_dir
CHECK_NAME=Stateless tests (debug, s3 storage)
REPO_COPY=${{runner.temp}}/stateless_s3_storage_debug/ClickHouse
KILL_TIMEOUT=10800
RUN_BY_HASH_NUM=0
RUN_BY_HASH_TOTAL=3
EOF
- name: Download json reports
uses: actions/download-artifact@v2
with:
path: ${{ env.REPORTS_PATH }}
- name: Clear repository
run: |
sudo rm -fr "$GITHUB_WORKSPACE" && mkdir "$GITHUB_WORKSPACE"
- name: Check out repository code
uses: actions/checkout@v2
- name: Functional test
run: |
sudo rm -fr "$TEMP_PATH"
mkdir -p "$TEMP_PATH"
cp -r "$GITHUB_WORKSPACE" "$TEMP_PATH"
cd "$REPO_COPY/tests/ci"
python3 functional_test_check.py "$CHECK_NAME" "$KILL_TIMEOUT"
- name: Cleanup
if: always()
run: |
docker kill "$(docker ps -q)" ||:
docker rm -f "$(docker ps -a -q)" ||:
sudo rm -fr "$TEMP_PATH"
FunctionalStatelessTestS3Debug1:
needs: [BuilderDebDebug]
runs-on: [self-hosted, func-tester]
steps:
- name: Set envs
run: |
cat >> "$GITHUB_ENV" << 'EOF'
TEMP_PATH=${{runner.temp}}/stateless_s3_storage_debug
REPORTS_PATH=${{runner.temp}}/reports_dir
CHECK_NAME=Stateless tests (debug, s3 storage)
REPO_COPY=${{runner.temp}}/stateless_s3_storage_debug/ClickHouse
KILL_TIMEOUT=10800
RUN_BY_HASH_NUM=1
RUN_BY_HASH_TOTAL=3
EOF
- name: Download json reports
uses: actions/download-artifact@v2
with:
path: ${{ env.REPORTS_PATH }}
- name: Clear repository
run: |
sudo rm -fr "$GITHUB_WORKSPACE" && mkdir "$GITHUB_WORKSPACE"
- name: Check out repository code
uses: actions/checkout@v2
- name: Functional test
run: |
sudo rm -fr "$TEMP_PATH"
mkdir -p "$TEMP_PATH"
cp -r "$GITHUB_WORKSPACE" "$TEMP_PATH"
cd "$REPO_COPY/tests/ci"
python3 functional_test_check.py "$CHECK_NAME" "$KILL_TIMEOUT"
- name: Cleanup
if: always()
run: |
docker kill "$(docker ps -q)" ||:
docker rm -f "$(docker ps -a -q)" ||:
sudo rm -fr "$TEMP_PATH"
FunctionalStatelessTestS3Debug2:
needs: [BuilderDebDebug]
runs-on: [self-hosted, func-tester]
steps:
- name: Set envs
run: |
cat >> "$GITHUB_ENV" << 'EOF'
TEMP_PATH=${{runner.temp}}/stateless_s3_storage_debug
REPORTS_PATH=${{runner.temp}}/reports_dir
CHECK_NAME=Stateless tests (debug, s3 storage)
REPO_COPY=${{runner.temp}}/stateless_s3_storage_debug/ClickHouse
KILL_TIMEOUT=10800
RUN_BY_HASH_NUM=2
RUN_BY_HASH_TOTAL=3
EOF
- name: Download json reports
uses: actions/download-artifact@v2
with:
path: ${{ env.REPORTS_PATH }}
- name: Clear repository
run: |
sudo rm -fr "$GITHUB_WORKSPACE" && mkdir "$GITHUB_WORKSPACE"
- name: Check out repository code
uses: actions/checkout@v2
- name: Functional test
run: |
sudo rm -fr "$TEMP_PATH"
mkdir -p "$TEMP_PATH"
cp -r "$GITHUB_WORKSPACE" "$TEMP_PATH"
cd "$REPO_COPY/tests/ci"
python3 functional_test_check.py "$CHECK_NAME" "$KILL_TIMEOUT"
- name: Cleanup
if: always()
run: |
docker kill "$(docker ps -q)" ||:
docker rm -f "$(docker ps -a -q)" ||:
sudo rm -fr "$TEMP_PATH"
FunctionalStatelessTestS3Tsan0:
needs: [BuilderDebTsan]
runs-on: [self-hosted, func-tester]
steps:
- name: Set envs
run: |
cat >> "$GITHUB_ENV" << 'EOF'
TEMP_PATH=${{runner.temp}}/stateless_s3_storage_tsan
REPORTS_PATH=${{runner.temp}}/reports_dir
CHECK_NAME=Stateless tests (tsan, s3 storage)
REPO_COPY=${{runner.temp}}/stateless_s3_storage_tsan/ClickHouse
KILL_TIMEOUT=10800
RUN_BY_HASH_NUM=0
RUN_BY_HASH_TOTAL=3
EOF
- name: Download json reports
uses: actions/download-artifact@v2
with:
path: ${{ env.REPORTS_PATH }}
- name: Clear repository
run: |
sudo rm -fr "$GITHUB_WORKSPACE" && mkdir "$GITHUB_WORKSPACE"
- name: Check out repository code
uses: actions/checkout@v2
- name: Functional test
run: |
sudo rm -fr "$TEMP_PATH"
mkdir -p "$TEMP_PATH"
cp -r "$GITHUB_WORKSPACE" "$TEMP_PATH"
cd "$REPO_COPY/tests/ci"
python3 functional_test_check.py "$CHECK_NAME" "$KILL_TIMEOUT"
- name: Cleanup
if: always()
run: |
docker kill "$(docker ps -q)" ||:
docker rm -f "$(docker ps -a -q)" ||:
sudo rm -fr "$TEMP_PATH"
FunctionalStatelessTestS3Tsan1:
needs: [BuilderDebTsan]
runs-on: [self-hosted, func-tester]
steps:
- name: Set envs
run: |
cat >> "$GITHUB_ENV" << 'EOF'
TEMP_PATH=${{runner.temp}}/stateless_s3_storage_tsan
REPORTS_PATH=${{runner.temp}}/reports_dir
CHECK_NAME=Stateless tests (tsan, s3 storage)
REPO_COPY=${{runner.temp}}/stateless_s3_storage_tsan/ClickHouse
KILL_TIMEOUT=10800
RUN_BY_HASH_NUM=1
RUN_BY_HASH_TOTAL=3
EOF
- name: Download json reports
uses: actions/download-artifact@v2
with:
path: ${{ env.REPORTS_PATH }}
- name: Clear repository
run: |
sudo rm -fr "$GITHUB_WORKSPACE" && mkdir "$GITHUB_WORKSPACE"
- name: Check out repository code
uses: actions/checkout@v2
- name: Functional test
run: |
sudo rm -fr "$TEMP_PATH"
mkdir -p "$TEMP_PATH"
cp -r "$GITHUB_WORKSPACE" "$TEMP_PATH"
cd "$REPO_COPY/tests/ci"
python3 functional_test_check.py "$CHECK_NAME" "$KILL_TIMEOUT"
- name: Cleanup
if: always()
run: |
docker kill "$(docker ps -q)" ||:
docker rm -f "$(docker ps -a -q)" ||:
sudo rm -fr "$TEMP_PATH"
FunctionalStatelessTestS3Tsan2:
needs: [BuilderDebTsan]
runs-on: [self-hosted, func-tester]
steps:
- name: Set envs
run: |
cat >> "$GITHUB_ENV" << 'EOF'
TEMP_PATH=${{runner.temp}}/stateless_s3_storage_tsan
REPORTS_PATH=${{runner.temp}}/reports_dir
CHECK_NAME=Stateless tests (tsan, s3 storage)
REPO_COPY=${{runner.temp}}/stateless_s3_storage_tsan/ClickHouse
KILL_TIMEOUT=10800
RUN_BY_HASH_NUM=2
RUN_BY_HASH_TOTAL=3
EOF
- name: Download json reports
uses: actions/download-artifact@v2
with:
path: ${{ env.REPORTS_PATH }}
- name: Clear repository
run: |
sudo rm -fr "$GITHUB_WORKSPACE" && mkdir "$GITHUB_WORKSPACE"
- name: Check out repository code
uses: actions/checkout@v2
- name: Functional test
run: |
sudo rm -fr "$TEMP_PATH"
mkdir -p "$TEMP_PATH"
cp -r "$GITHUB_WORKSPACE" "$TEMP_PATH"
cd "$REPO_COPY/tests/ci"
python3 functional_test_check.py "$CHECK_NAME" "$KILL_TIMEOUT"
- name: Cleanup
if: always()
run: |
docker kill "$(docker ps -q)" ||:
docker rm -f "$(docker ps -a -q)" ||:
sudo rm -fr "$TEMP_PATH"
FunctionalStatelessTestAarch64:
needs: [BuilderDebAarch64]
runs-on: [self-hosted, func-tester-aarch64]
@ -3388,6 +3610,12 @@ jobs:
- FunctionalStatefulTestMsan
- FunctionalStatefulTestUBsan
- FunctionalStatelessTestReleaseS3
- FunctionalStatelessTestS3Debug0
- FunctionalStatelessTestS3Debug1
- FunctionalStatelessTestS3Debug2
- FunctionalStatelessTestS3Tsan0
- FunctionalStatelessTestS3Tsan1
- FunctionalStatelessTestS3Tsan2
- StressTestDebug
- StressTestAsan
- StressTestTsan

2
CMakeLists.txt
View File

@ -143,6 +143,8 @@ include (cmake/add_warning.cmake)
if (COMPILER_CLANG)
# generate ranges for fast "addr2line" search
if (NOT CMAKE_BUILD_TYPE_UC STREQUAL "RELEASE")
# NOTE: that clang has a bug because of it does not emit .debug_aranges
# with ThinLTO, so custom ld.lld wrapper is shipped in docker images.
set(COMPILER_FLAGS "${COMPILER_FLAGS} -gdwarf-aranges")
endif ()

3
README.md
View File

@ -15,4 +15,5 @@ ClickHouse® is an open-source column-oriented database management system that a
* [Contacts](https://clickhouse.com/company/contact) can help to get your questions answered if there are any.
## Upcoming events
* [**v22.8 Release Webinar**](https://clickhouse.com/company/events/v22-8-release-webinar) Original creator, co-founder, and CTO of ClickHouse Alexey Milovidov will walk us through the highlights of the release, provide live demos, and share vision into what is coming in the roadmap.
* [**v22.9 Release Webinar**](https://clickhouse.com/company/events/v22-9-release-webinar) Original creator, co-founder, and CTO of ClickHouse Alexey Milovidov will walk us through the highlights of the release, provide live demos, and share vision into what is coming in the roadmap.
* [**ClickHouse for Analytics @ Barracuda Networks**](https://www.meetup.com/clickhouse-silicon-valley-meetup-group/events/288140358/) Join us for this in person meetup hosted by our friends at Barracuda in Bay Area.

1
base/glibc-compatibility/memcpy/memcpy.h
View File

@ -214,4 +214,3 @@ tail:
return ret;
}

17
cmake/ld.lld.in Executable file
View File

@ -0,0 +1,17 @@
#!/usr/bin/env bash
# This is a workaround for bug in llvm/clang,
# that does not produce .debug_aranges with LTO
#
# NOTE: this is a temporary solution, that should be removed once [1] will be
# resolved.
#
# [1]: https://discourse.llvm.org/t/clang-does-not-produce-full-debug-aranges-section-with-thinlto/64898/8
# NOTE: only -flto=thin is supported.
# NOTE: it is not possible to check was there -gdwarf-aranges initially or not.
if [[ "$*" =~ -plugin-opt=thinlto ]]; then
exec "@LLD_PATH@" -mllvm -generate-arange-section "$@"
else
exec "@LLD_PATH@" "$@"
fi

2
cmake/split_debug_symbols.cmake
View File

@ -20,7 +20,7 @@ macro(clickhouse_split_debug_symbols)
COMMAND mkdir -p "${STRIP_DESTINATION_DIR}/bin"
COMMAND cp "${STRIP_BINARY_PATH}" "${STRIP_DESTINATION_DIR}/bin/${STRIP_TARGET}"
# Splits debug symbols into separate file, leaves the binary untouched:
COMMAND "${OBJCOPY_PATH}" --only-keep-debug --compress-debug-sections "${STRIP_DESTINATION_DIR}/bin/${STRIP_TARGET}" "${STRIP_DESTINATION_DIR}/lib/debug/bin/${STRIP_TARGET}.debug"
COMMAND "${OBJCOPY_PATH}" --only-keep-debug "${STRIP_DESTINATION_DIR}/bin/${STRIP_TARGET}" "${STRIP_DESTINATION_DIR}/lib/debug/bin/${STRIP_TARGET}.debug"
COMMAND chmod 0644 "${STRIP_DESTINATION_DIR}/lib/debug/bin/${STRIP_TARGET}.debug"
# Strips binary, sections '.note' & '.comment' are removed in line with Debian's stripping policy: www.debian.org/doc/debian-policy/ch-files.html, section '.clickhouse.hash' is needed for integrity check:
COMMAND "${STRIP_PATH}" --remove-section=.comment --remove-section=.note --keep-section=.clickhouse.hash "${STRIP_DESTINATION_DIR}/bin/${STRIP_TARGET}"

9
cmake/tools.cmake
View File

@ -94,8 +94,13 @@ if (LINKER_NAME)
if (NOT LLD_PATH)
message (FATAL_ERROR "Using linker ${LINKER_NAME} but can't find its path.")
endif ()
set (CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} --ld-path=${LLD_PATH}")
set (CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS} --ld-path=${LLD_PATH}")
# This a temporary quirk to emit .debug_aranges with ThinLTO
set (LLD_WRAPPER "${CMAKE_CURRENT_BINARY_DIR}/ld.lld")
configure_file ("${CMAKE_CURRENT_SOURCE_DIR}/cmake/ld.lld.in" "${LLD_WRAPPER}" @ONLY)
set (CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} --ld-path=${LLD_WRAPPER}")
set (CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS} --ld-path=${LLD_WRAPPER}")
else ()
set (CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -fuse-ld=${LINKER_NAME}")
set (CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS} -fuse-ld=${LINKER_NAME}")

2
contrib/NuRaft vendored

@ -1 +1 @@
Subproject commit bdba298189e29995892de78dcecf64d127444e81
Subproject commit 1be805e7cb2494aa8170015493474379b0362dfc

2
contrib/datasketches-cpp vendored

@ -1 +1 @@
Subproject commit 7d73d7610db31d4e1ecde0fb3a7ee90ef371207f
Subproject commit 7abd49bb2e72bf9a5029993d31dcb1872da88292

5
contrib/libcxx-cmake/CMakeLists.txt
View File

@ -54,9 +54,8 @@ set(SRCS
add_library(cxx ${SRCS})
set_target_properties(cxx PROPERTIES FOLDER "contrib/libcxx-cmake")
target_include_directories(cxx SYSTEM BEFORE PUBLIC
$<BUILD_INTERFACE:${LIBCXX_SOURCE_DIR}/include>
$<BUILD_INTERFACE:${LIBCXX_SOURCE_DIR}>/src)
target_include_directories(cxx SYSTEM BEFORE PRIVATE $<BUILD_INTERFACE:${LIBCXX_SOURCE_DIR}/src>)
target_include_directories(cxx SYSTEM BEFORE PUBLIC $<BUILD_INTERFACE:${LIBCXX_SOURCE_DIR}/include>)
target_compile_definitions(cxx PRIVATE -D_LIBCPP_BUILDING_LIBRARY -DLIBCXX_BUILDING_LIBCXXABI)
# Enable capturing stack traces for all exceptions.

3
docker/packager/binary/Dockerfile
View File

@ -83,5 +83,8 @@ RUN export CODENAME="$(lsb_release --codename --short | tr 'A-Z' 'a-z')" \
--yes --no-install-recommends \
&& apt-get clean
# for external_symbolizer_path
RUN ln -s /usr/bin/llvm-symbolizer-15 /usr/bin/llvm-symbolizer
COPY build.sh /
CMD ["bash", "-c", "/build.sh 2>&1"]

3
docker/server/Dockerfile.ubuntu
View File

@ -31,9 +31,6 @@ ARG deb_location_url=""
# set non-empty single_binary_location_url to create docker image
# from a single binary url (useful for non-standard builds - with sanitizers, for arm64).
# for example (run on aarch64 server):
# docker build . --network host --build-arg single_binary_location_url="https://builds.clickhouse.com/master/aarch64/clickhouse" -t altinity/clickhouse-server:master-testing-arm
# note: clickhouse-odbc-bridge is not supported there.
ARG single_binary_location_url=""
# user/group precreated explicitly with fixed uid/gid on purpose.

2
docker/server/entrypoint.sh
View File

@ -37,7 +37,6 @@ if [ -n "$ERROR_LOG_PATH" ]; then ERROR_LOG_DIR="$(dirname "$ERROR_LOG_PATH")";
FORMAT_SCHEMA_PATH="$(clickhouse extract-from-config --config-file "$CLICKHOUSE_CONFIG" --key=format_schema_path || true)"
# There could be many disks declared in config
readarray -t FILESYSTEM_CACHE_PATHS < <(clickhouse extract-from-config --config-file "$CLICKHOUSE_CONFIG" --key='storage_configuration.disks.*.data_cache_path' || true)
readarray -t DISKS_PATHS < <(clickhouse extract-from-config --config-file "$CLICKHOUSE_CONFIG" --key='storage_configuration.disks.*.path' || true)
CLICKHOUSE_USER="${CLICKHOUSE_USER:-default}"
@ -51,7 +50,6 @@ for dir in "$DATA_DIR" \
"$TMP_DIR" \
"$USER_PATH" \
"$FORMAT_SCHEMA_PATH" \
"${FILESYSTEM_CACHE_PATHS[@]}" \
"${DISKS_PATHS[@]}"
do
# check if variable not empty

49
docker/test/stress/run.sh
View File

@ -3,6 +3,9 @@
# shellcheck disable=SC2086
# shellcheck disable=SC2024
# Avoid overlaps with previous runs
dmesg --clear
set -x
# Thread Fuzzer allows to check more permutations of possible thread scheduling
@ -38,8 +41,10 @@ function install_packages()
function configure()
{
export ZOOKEEPER_FAULT_INJECTION=1
# install test configs
export USE_DATABASE_ORDINARY=1
export EXPORT_S3_STORAGE_POLICIES=1
/usr/share/clickhouse-test/config/install.sh
# we mount tests folder from repo to /usr/share
@ -183,11 +188,11 @@ install_packages package_folder
configure
azurite-blob --blobHost 0.0.0.0 --blobPort 10000 --debug /azurite_log &
./setup_minio.sh stateful # to have a proper environment
./setup_minio.sh stateless # to have a proper environment
start
# shellcheck disable=SC2086 # No quotes because I want to split it into words.
shellcheck disable=SC2086 # No quotes because I want to split it into words.
/s3downloader --url-prefix "$S3_URL" --dataset-names $DATASETS
chmod 777 -R /var/lib/clickhouse
clickhouse-client --query "ATTACH DATABASE IF NOT EXISTS datasets ENGINE = Ordinary"
@ -200,12 +205,36 @@ start
clickhouse-client --query "SHOW TABLES FROM datasets"
clickhouse-client --query "SHOW TABLES FROM test"
clickhouse-client --query "RENAME TABLE datasets.hits_v1 TO test.hits"
clickhouse-client --query "RENAME TABLE datasets.visits_v1 TO test.visits"
clickhouse-client --query "CREATE TABLE test.hits_s3 (WatchID UInt64, JavaEnable UInt8, Title String, GoodEvent Int16, EventTime DateTime, EventDate Date, CounterID UInt32, ClientIP UInt32, ClientIP6 FixedString(16), RegionID UInt32, UserID UInt64, CounterClass Int8, OS UInt8, UserAgent UInt8, URL String, Referer String, URLDomain String, RefererDomain String, Refresh UInt8, IsRobot UInt8, RefererCategories Array(UInt16), URLCategories Array(UInt16), URLRegions Array(UInt32), RefererRegions Array(UInt32), ResolutionWidth UInt16, ResolutionHeight UInt16, ResolutionDepth UInt8, FlashMajor UInt8, FlashMinor UInt8, FlashMinor2 String, NetMajor UInt8, NetMinor UInt8, UserAgentMajor UInt16, UserAgentMinor FixedString(2), CookieEnable UInt8, JavascriptEnable UInt8, IsMobile UInt8, MobilePhone UInt8, MobilePhoneModel String, Params String, IPNetworkID UInt32, TraficSourceID Int8, SearchEngineID UInt16, SearchPhrase String, AdvEngineID UInt8, IsArtifical UInt8, WindowClientWidth UInt16, WindowClientHeight UInt16, ClientTimeZone Int16, ClientEventTime DateTime, SilverlightVersion1 UInt8, SilverlightVersion2 UInt8, SilverlightVersion3 UInt32, SilverlightVersion4 UInt16, PageCharset String, CodeVersion UInt32, IsLink UInt8, IsDownload UInt8, IsNotBounce UInt8, FUniqID UInt64, HID UInt32, IsOldCounter UInt8, IsEvent UInt8, IsParameter UInt8, DontCountHits UInt8, WithHash UInt8, HitColor FixedString(1), UTCEventTime DateTime, Age UInt8, Sex UInt8, Income UInt8, Interests UInt16, Robotness UInt8, GeneralInterests Array(UInt16), RemoteIP UInt32, RemoteIP6 FixedString(16), WindowName Int32, OpenerName Int32, HistoryLength Int16, BrowserLanguage FixedString(2), BrowserCountry FixedString(2), SocialNetwork String, SocialAction String, HTTPError UInt16, SendTiming Int32, DNSTiming Int32, ConnectTiming Int32, ResponseStartTiming Int32, ResponseEndTiming Int32, FetchTiming Int32, RedirectTiming Int32, DOMInteractiveTiming Int32, DOMContentLoadedTiming Int32, DOMCompleteTiming Int32, LoadEventStartTiming Int32, LoadEventEndTiming Int32, NSToDOMContentLoadedTiming Int32, FirstPaintTiming Int32, RedirectCount Int8, SocialSourceNetworkID UInt8, SocialSourcePage String, ParamPrice Int64, ParamOrderID String, ParamCurrency FixedString(3), ParamCurrencyID UInt16, GoalsReached Array(UInt32), OpenstatServiceName String, OpenstatCampaignID String, OpenstatAdID String, OpenstatSourceID String, UTMSource String, UTMMedium String, UTMCampaign String, UTMContent String, UTMTerm String, FromTag String, HasGCLID UInt8, RefererHash UInt64, URLHash UInt64, CLID UInt32, YCLID UInt64, ShareService String, ShareURL String, ShareTitle String, ParsedParams Nested(Key1 String, Key2 String, Key3 String, Key4 String, Key5 String, ValueDouble Float64), IslandID FixedString(16), RequestNum UInt32, RequestTry UInt8) ENGINE = MergeTree() PARTITION BY toYYYYMM(EventDate) ORDER BY (CounterID, EventDate, intHash32(UserID)) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192, storage_policy='s3_cache'"
clickhouse-client --query "INSERT INTO test.hits_s3 SELECT * FROM test.hits"
clickhouse-client --query "CREATE TABLE test.hits_s3 (WatchID UInt64, JavaEnable UInt8, Title String, GoodEvent Int16, EventTime DateTime, EventDate Date, CounterID UInt32, ClientIP UInt32, ClientIP6 FixedString(16), RegionID UInt32, UserID UInt64, CounterClass Int8, OS UInt8, UserAgent UInt8, URL String, Referer String, URLDomain String, RefererDomain String, Refresh UInt8, IsRobot UInt8, RefererCategories Array(UInt16), URLCategories Array(UInt16), URLRegions Array(UInt32), RefererRegions Array(UInt32), ResolutionWidth UInt16, ResolutionHeight UInt16, ResolutionDepth UInt8, FlashMajor UInt8, FlashMinor UInt8, FlashMinor2 String, NetMajor UInt8, NetMinor UInt8, UserAgentMajor UInt16, UserAgentMinor FixedString(2), CookieEnable UInt8, JavascriptEnable UInt8, IsMobile UInt8, MobilePhone UInt8, MobilePhoneModel String, Params String, IPNetworkID UInt32, TraficSourceID Int8, SearchEngineID UInt16, SearchPhrase String, AdvEngineID UInt8, IsArtifical UInt8, WindowClientWidth UInt16, WindowClientHeight UInt16, ClientTimeZone Int16, ClientEventTime DateTime, SilverlightVersion1 UInt8, SilverlightVersion2 UInt8, SilverlightVersion3 UInt32, SilverlightVersion4 UInt16, PageCharset String, CodeVersion UInt32, IsLink UInt8, IsDownload UInt8, IsNotBounce UInt8, FUniqID UInt64, HID UInt32, IsOldCounter UInt8, IsEvent UInt8, IsParameter UInt8, DontCountHits UInt8, WithHash UInt8, HitColor FixedString(1), UTCEventTime DateTime, Age UInt8, Sex UInt8, Income UInt8, Interests UInt16, Robotness UInt8, GeneralInterests Array(UInt16), RemoteIP UInt32, RemoteIP6 FixedString(16), WindowName Int32, OpenerName Int32, HistoryLength Int16, BrowserLanguage FixedString(2), BrowserCountry FixedString(2), SocialNetwork String, SocialAction String, HTTPError UInt16, SendTiming Int32, DNSTiming Int32, ConnectTiming Int32, ResponseStartTiming Int32, ResponseEndTiming Int32, FetchTiming Int32, RedirectTiming Int32, DOMInteractiveTiming Int32, DOMContentLoadedTiming Int32, DOMCompleteTiming Int32, LoadEventStartTiming Int32, LoadEventEndTiming Int32, NSToDOMContentLoadedTiming Int32, FirstPaintTiming Int32, RedirectCount Int8, SocialSourceNetworkID UInt8, SocialSourcePage String, ParamPrice Int64, ParamOrderID String, ParamCurrency FixedString(3), ParamCurrencyID UInt16, GoalsReached Array(UInt32), OpenstatServiceName String, OpenstatCampaignID String, OpenstatAdID String, OpenstatSourceID String, UTMSource String, UTMMedium String, UTMCampaign String, UTMContent String, UTMTerm String, FromTag String, HasGCLID UInt8, RefererHash UInt64, URLHash UInt64, CLID UInt32, YCLID UInt64, ShareService String, ShareURL String, ShareTitle String, ParsedParams Nested(Key1 String, Key2 String, Key3 String, Key4 String, Key5 String, ValueDouble Float64), IslandID FixedString(16), RequestNum UInt32, RequestTry UInt8) ENGINE = MergeTree() PARTITION BY toYYYYMM(EventDate) ORDER BY (CounterID, EventDate, intHash32(UserID)) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192, storage_policy='s3_cache'"
clickhouse-client --query "CREATE TABLE test.hits (WatchID UInt64, JavaEnable UInt8, Title String, GoodEvent Int16, EventTime DateTime, EventDate Date, CounterID UInt32, ClientIP UInt32, ClientIP6 FixedString(16), RegionID UInt32, UserID UInt64, CounterClass Int8, OS UInt8, UserAgent UInt8, URL String, Referer String, URLDomain String, RefererDomain String, Refresh UInt8, IsRobot UInt8, RefererCategories Array(UInt16), URLCategories Array(UInt16), URLRegions Array(UInt32), RefererRegions Array(UInt32), ResolutionWidth UInt16, ResolutionHeight UInt16, ResolutionDepth UInt8, FlashMajor UInt8, FlashMinor UInt8, FlashMinor2 String, NetMajor UInt8, NetMinor UInt8, UserAgentMajor UInt16, UserAgentMinor FixedString(2), CookieEnable UInt8, JavascriptEnable UInt8, IsMobile UInt8, MobilePhone UInt8, MobilePhoneModel String, Params String, IPNetworkID UInt32, TraficSourceID Int8, SearchEngineID UInt16, SearchPhrase String, AdvEngineID UInt8, IsArtifical UInt8, WindowClientWidth UInt16, WindowClientHeight UInt16, ClientTimeZone Int16, ClientEventTime DateTime, SilverlightVersion1 UInt8, SilverlightVersion2 UInt8, SilverlightVersion3 UInt32, SilverlightVersion4 UInt16, PageCharset String, CodeVersion UInt32, IsLink UInt8, IsDownload UInt8, IsNotBounce UInt8, FUniqID UInt64, HID UInt32, IsOldCounter UInt8, IsEvent UInt8, IsParameter UInt8, DontCountHits UInt8, WithHash UInt8, HitColor FixedString(1), UTCEventTime DateTime, Age UInt8, Sex UInt8, Income UInt8, Interests UInt16, Robotness UInt8, GeneralInterests Array(UInt16), RemoteIP UInt32, RemoteIP6 FixedString(16), WindowName Int32, OpenerName Int32, HistoryLength Int16, BrowserLanguage FixedString(2), BrowserCountry FixedString(2), SocialNetwork String, SocialAction String, HTTPError UInt16, SendTiming Int32, DNSTiming Int32, ConnectTiming Int32, ResponseStartTiming Int32, ResponseEndTiming Int32, FetchTiming Int32, RedirectTiming Int32, DOMInteractiveTiming Int32, DOMContentLoadedTiming Int32, DOMCompleteTiming Int32, LoadEventStartTiming Int32, LoadEventEndTiming Int32, NSToDOMContentLoadedTiming Int32, FirstPaintTiming Int32, RedirectCount Int8, SocialSourceNetworkID UInt8, SocialSourcePage String, ParamPrice Int64, ParamOrderID String, ParamCurrency FixedString(3), ParamCurrencyID UInt16, GoalsReached Array(UInt32), OpenstatServiceName String, OpenstatCampaignID String, OpenstatAdID String, OpenstatSourceID String, UTMSource String, UTMMedium String, UTMCampaign String, UTMContent String, UTMTerm String, FromTag String, HasGCLID UInt8, RefererHash UInt64, URLHash UInt64, CLID UInt32, YCLID UInt64, ShareService String, ShareURL String, ShareTitle String, ParsedParams Nested(Key1 String, Key2 String, Key3 String, Key4 String, Key5 String, ValueDouble Float64), IslandID FixedString(16), RequestNum UInt32, RequestTry UInt8) ENGINE = MergeTree() PARTITION BY toYYYYMM(EventDate) ORDER BY (CounterID, EventDate, intHash32(UserID)) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192, storage_policy='s3_cache'"
clickhouse-client --query "CREATE TABLE test.visits (CounterID UInt32, StartDate Date, Sign Int8, IsNew UInt8, VisitID UInt64, UserID UInt64, StartTime DateTime, Duration UInt32, UTCStartTime DateTime, PageViews Int32, Hits Int32, IsBounce UInt8, Referer String, StartURL String, RefererDomain String, StartURLDomain String, EndURL String, LinkURL String, IsDownload UInt8, TraficSourceID Int8, SearchEngineID UInt16, SearchPhrase String, AdvEngineID UInt8, PlaceID Int32, RefererCategories Array(UInt16), URLCategories Array(UInt16), URLRegions Array(UInt32), RefererRegions Array(UInt32), IsYandex UInt8, GoalReachesDepth Int32, GoalReachesURL Int32, GoalReachesAny Int32, SocialSourceNetworkID UInt8, SocialSourcePage String, MobilePhoneModel String, ClientEventTime DateTime, RegionID UInt32, ClientIP UInt32, ClientIP6 FixedString(16), RemoteIP UInt32, RemoteIP6 FixedString(16), IPNetworkID UInt32, SilverlightVersion3 UInt32, CodeVersion UInt32, ResolutionWidth UInt16, ResolutionHeight UInt16, UserAgentMajor UInt16, UserAgentMinor UInt16, WindowClientWidth UInt16, WindowClientHeight UInt16, SilverlightVersion2 UInt8, SilverlightVersion4 UInt16, FlashVersion3 UInt16, FlashVersion4 UInt16, ClientTimeZone Int16, OS UInt8, UserAgent UInt8, ResolutionDepth UInt8, FlashMajor UInt8, FlashMinor UInt8, NetMajor UInt8, NetMinor UInt8, MobilePhone UInt8, SilverlightVersion1 UInt8, Age UInt8, Sex UInt8, Income UInt8, JavaEnable UInt8, CookieEnable UInt8, JavascriptEnable UInt8, IsMobile UInt8, BrowserLanguage UInt16, BrowserCountry UInt16, Interests UInt16, Robotness UInt8, GeneralInterests Array(UInt16), Params Array(String), Goals Nested(ID UInt32, Serial UInt32, EventTime DateTime, Price Int64, OrderID String, CurrencyID UInt32), WatchIDs Array(UInt64), ParamSumPrice Int64, ParamCurrency FixedString(3), ParamCurrencyID UInt16, ClickLogID UInt64, ClickEventID Int32, ClickGoodEvent Int32, ClickEventTime DateTime, ClickPriorityID Int32, ClickPhraseID Int32, ClickPageID Int32, ClickPlaceID Int32, ClickTypeID Int32, ClickResourceID Int32, ClickCost UInt32, ClickClientIP UInt32, ClickDomainID UInt32, ClickURL String, ClickAttempt UInt8, ClickOrderID UInt32, ClickBannerID UInt32, ClickMarketCategoryID UInt32, ClickMarketPP UInt32, ClickMarketCategoryName String, ClickMarketPPName String, ClickAWAPSCampaignName String, ClickPageName String, ClickTargetType UInt16, ClickTargetPhraseID UInt64, ClickContextType UInt8, ClickSelectType Int8, ClickOptions String, ClickGroupBannerID Int32, OpenstatServiceName String, OpenstatCampaignID String, OpenstatAdID String, OpenstatSourceID String, UTMSource String, UTMMedium String, UTMCampaign String, UTMContent String, UTMTerm String, FromTag String, HasGCLID UInt8, FirstVisit DateTime, PredLastVisit Date, LastVisit Date, TotalVisits UInt32, TraficSource Nested(ID Int8, SearchEngineID UInt16, AdvEngineID UInt8, PlaceID UInt16, SocialSourceNetworkID UInt8, Domain String, SearchPhrase String, SocialSourcePage String), Attendance FixedString(16), CLID UInt32, YCLID UInt64, NormalizedRefererHash UInt64, SearchPhraseHash UInt64, RefererDomainHash UInt64, NormalizedStartURLHash UInt64, StartURLDomainHash UInt64, NormalizedEndURLHash UInt64, TopLevelDomain UInt64, URLScheme UInt64, OpenstatServiceNameHash UInt64, OpenstatCampaignIDHash UInt64, OpenstatAdIDHash UInt64, OpenstatSourceIDHash UInt64, UTMSourceHash UInt64, UTMMediumHash UInt64, UTMCampaignHash UInt64, UTMContentHash UInt64, UTMTermHash UInt64, FromHash UInt64, WebVisorEnabled UInt8, WebVisorActivity UInt32, ParsedParams Nested(Key1 String, Key2 String, Key3 String, Key4 String, Key5 String, ValueDouble Float64), Market Nested(Type UInt8, GoalID UInt32, OrderID String, OrderPrice Int64, PP UInt32, DirectPlaceID UInt32, DirectOrderID UInt32, DirectBannerID UInt32, GoodID String, GoodName String, GoodQuantity Int32, GoodPrice Int64), IslandID FixedString(16)) ENGINE = CollapsingMergeTree(Sign) PARTITION BY toYYYYMM(StartDate) ORDER BY (CounterID, StartDate, intHash32(UserID), VisitID) SAMPLE BY intHash32(UserID) SETTINGS index_granularity = 8192, storage_policy='s3_cache'"
clickhouse-client --query "INSERT INTO test.hits_s3 SELECT * FROM datasets.hits_v1 SETTINGS enable_filesystem_cache_on_write_operations=0"
clickhouse-client --query "INSERT INTO test.hits SELECT * FROM datasets.hits_v1 SETTINGS enable_filesystem_cache_on_write_operations=0"
clickhouse-client --query "INSERT INTO test.visits SELECT * FROM datasets.visits_v1 SETTINGS enable_filesystem_cache_on_write_operations=0"
clickhouse-client --query "DROP TABLE datasets.visits_v1 SYNC"
clickhouse-client --query "DROP TABLE datasets.hits_v1 SYNC"
clickhouse-client --query "SHOW TABLES FROM test"
clickhouse-client --query "SYSTEM STOP THREAD FUZZER"
stop
# Let's enable S3 storage by default
export USE_S3_STORAGE_FOR_MERGE_TREE=1
configure
# But we still need default disk because some tables loaded only into it
sudo cat /etc/clickhouse-server/config.d/s3_storage_policy_by_default.xml | sed "s|<disk>s3</disk>|<disk>s3</disk><disk>default</disk>|" > /etc/clickhouse-server/config.d/s3_storage_policy_by_default.xml.tmp
mv /etc/clickhouse-server/config.d/s3_storage_policy_by_default.xml.tmp /etc/clickhouse-server/config.d/s3_storage_policy_by_default.xml
sudo chown clickhouse /etc/clickhouse-server/config.d/s3_storage_policy_by_default.xml
sudo chgrp clickhouse /etc/clickhouse-server/config.d/s3_storage_policy_by_default.xml
start
./stress --hung-check --drop-databases --output-folder test_output --skip-func-tests "$SKIP_TESTS_OPTION" \
&& echo -e 'Test script exit code\tOK' >> /test_output/test_results.tsv \
|| echo -e 'Test script failed\tFAIL' >> /test_output/test_results.tsv
@ -255,6 +284,14 @@ zgrep -Fa "Code: 49, e.displayText() = DB::Exception:" /var/log/clickhouse-serve
# Remove file logical_errors.txt if it's empty
[ -s /test_output/logical_errors.txt ] || rm /test_output/logical_errors.txt
# No such key errors
zgrep -Ea "Code: 499.*The specified key does not exist" /var/log/clickhouse-server/clickhouse-server*.log > /test_output/no_such_key_errors.txt \
&& echo -e 'S3_ERROR No such key thrown (see clickhouse-server.log or no_such_key_errors.txt)\tFAIL' >> /test_output/test_results.tsv \
|| echo -e 'No lost s3 keys\tOK' >> /test_output/test_results.tsv
# Remove file no_such_key_errors.txt if it's empty
[ -s /test_output/no_such_key_errors.txt ] || rm /test_output/no_such_key_errors.txt
# Crash
zgrep -Fa "########################################" /var/log/clickhouse-server/clickhouse-server*.log > /dev/null \
&& echo -e 'Killed by signal (in clickhouse-server.log)\tFAIL' >> /test_output/test_results.tsv \

2
docker/test/stress/stress
View File

@ -168,7 +168,7 @@ def prepare_for_hung_check(drop_databases):
for db in databases:
if db == "system":
continue
command = make_query_command(f"DROP DATABASE {db}")
command = make_query_command(f'DETACH DATABASE {db}')
# we don't wait for drop
Popen(command, shell=True)
break

2
docker/test/style/Dockerfile
View File

@ -17,7 +17,7 @@ RUN apt-get update && env DEBIAN_FRONTEND=noninteractive apt-get install --yes \
python3-pip \
shellcheck \
yamllint \
&& pip3 install black boto3 codespell dohq-artifactory PyGithub unidiff pylint==2.6.2 \
&& pip3 install black==22.8.0 boto3 codespell==2.2.1 dohq-artifactory PyGithub unidiff pylint==2.6.2 \
&& apt-get clean \
&& rm -rf /root/.cache/pip

2
docs/en/development/build-osx.md
View File

@ -37,7 +37,7 @@ sudo xcode-select --install
``` bash
brew update
brew install cmake ninja libtool gettext llvm gcc binutils grep findutils
brew install ccache cmake ninja libtool gettext llvm gcc binutils grep findutils
```
## Checkout ClickHouse Sources {#checkout-clickhouse-sources}

4
docs/en/engines/database-engines/replicated.md
View File

@ -12,7 +12,7 @@ One ClickHouse server can have multiple replicated databases running and updatin
## Creating a Database {#creating-a-database}
``` sql
CREATE DATABASE testdb ENGINE = Replicated('zoo_path', 'shard_name', 'replica_name') [SETTINGS ...]
CREATE DATABASE testdb ENGINE = Replicated('zoo_path', 'shard_name', 'replica_name') [SETTINGS ...]
```
**Engine Parameters**
@ -21,9 +21,7 @@ One ClickHouse server can have multiple replicated databases running and updatin
- `shard_name` — Shard name. Database replicas are grouped into shards by `shard_name`.
- `replica_name` — Replica name. Replica names must be different for all replicas of the same shard.
:::warning
For [ReplicatedMergeTree](../table-engines/mergetree-family/replication.md#table_engines-replication) tables if no arguments provided, then default arguments are used: `/clickhouse/tables/{uuid}/{shard}` and `{replica}`. These can be changed in the server settings [default_replica_path](../../operations/server-configuration-parameters/settings.md#default_replica_path) and [default_replica_name](../../operations/server-configuration-parameters/settings.md#default_replica_name). Macro `{uuid}` is unfolded to table's uuid, `{shard}` and `{replica}` are unfolded to values from server config, not from database engine arguments. But in the future, it will be possible to use `shard_name` and `replica_name` of Replicated database.
:::
## Specifics and Recommendations {#specifics-and-recommendations}

4
docs/en/engines/table-engines/integrations/embedded-rocksdb.md
View File

@ -16,12 +16,14 @@ CREATE TABLE [IF NOT EXISTS] [db.]table_name [ON CLUSTER cluster]
name1 [type1] [DEFAULT|MATERIALIZED|ALIAS expr1],
name2 [type2] [DEFAULT|MATERIALIZED|ALIAS expr2],
...
) ENGINE = EmbeddedRocksDB([ttl]) PRIMARY KEY(primary_key_name)
) ENGINE = EmbeddedRocksDB([ttl, rocksdb_dir, read_only]) PRIMARY KEY(primary_key_name)
```
Engine parameters:
- `ttl` - time to live for values. TTL is accepted in seconds. If TTL is 0, regular RocksDB instance is used (without TTL).
- `rocksdb_dir` - path to the directory of an existed RocksDB or the destination path of the created RocksDB. Open the table with the specified `rocksdb_dir`.
- `read_only` - when `read_only` is set to true, read-only mode is used. For storage with TTL, compaction will not be triggered (neither manual nor automatic), so no expired entries are removed.
- `primary_key_name` – any column name in the column list.
- `primary key` must be specified, it supports only one column in the primary key. The primary key will be serialized in binary as a `rocksdb key`.
- columns other than the primary key will be serialized in binary as `rocksdb` value in corresponding order.

2
docs/en/engines/table-engines/special/generate.md
View File

@ -15,7 +15,7 @@ Usage examples:
## Usage in ClickHouse Server {#usage-in-clickhouse-server}
``` sql
ENGINE = GenerateRandom(random_seed, max_string_length, max_array_length)
ENGINE = GenerateRandom([random_seed] [,max_string_length] [,max_array_length])
```
The `max_array_length` and `max_string_length` parameters specify maximum length of all

654
docs/en/getting-started/example-datasets/nypd_complaint_data.md Normal file
View File

@ -0,0 +1,654 @@
---
slug: /en/getting-started/example-datasets/nypd_complaint_data
sidebar_label: NYPD Complaint Data
description: "Ingest and query Tab Separated Value data in 5 steps"
title: NYPD Complaint Data
---
Tab separated value, or TSV, files are common and may include field headings as the first line of the file. ClickHouse can ingest TSVs, and also can query TSVs without ingesting the files. This guide covers both of these cases. If you need to query or ingest CSV files, the same techniques work, simply substitute `TSV` with `CSV` in your format arguments.
While working through this guide you will:
- **Investigate**: Query the structure and content of the TSV file.
- **Determine the target ClickHouse schema**: Choose proper data types and map the existing data to those types.
- **Create a ClickHouse table**.
- **Preprocess and stream** the data to ClickHouse.
- **Run some queries** against ClickHouse.
The dataset used in this guide comes from the NYC Open Data team, and contains data about "all valid felony, misdemeanor, and violation crimes reported to the New York City Police Department (NYPD)". At the time of writing, the data file is 166MB, but it is updated regularly.
**Source**: [data.cityofnewyork.us](https://data.cityofnewyork.us/Public-Safety/NYPD-Complaint-Data-Current-Year-To-Date-/5uac-w243)
**Terms of use**: https://www1.nyc.gov/home/terms-of-use.page
## Prerequisites
- Download the dataset by visiting the [NYPD Complaint Data Current (Year To Date)](https://data.cityofnewyork.us/Public-Safety/NYPD-Complaint-Data-Current-Year-To-Date-/5uac-w243) page, clicking the Export button, and choosing **TSV for Excel**.
- Install [ClickHouse server and client](../../getting-started/install.md).
- [Launch](../../getting-started/install.md#launch) ClickHouse server, and connect with `clickhouse-client`
### A note about the commands described in this guide
There are two types of commands in this guide:
- Some of the commands are querying the TSV files, these are run at the command prompt.
- The rest of the commands are querying ClickHouse, and these are run in the `clickhouse-client` or Play UI.
:::note
The examples in this guide assume that you have saved the TSV file to `${HOME}/NYPD_Complaint_Data_Current__Year_To_Date_.tsv`, please adjust the commands if needed.
:::
## Familiarize yourself with the TSV file
Before starting to work with the ClickHouse database familiarize yourself with the data.
### Look at the fields in the source TSV file
This is an example of a command to query a TSV file, but don't run it yet.
```sh
clickhouse-local --query \
"describe file('${HOME}/NYPD_Complaint_Data_Current__Year_To_Date_.tsv', 'TSVWithNames')"
```
Sample response
```response
CMPLNT_NUM Nullable(Float64)
ADDR_PCT_CD Nullable(Float64)
BORO_NM Nullable(String)
CMPLNT_FR_DT Nullable(String)
CMPLNT_FR_TM Nullable(String)
```
:::tip
Most of the time the above command will let you know which fields in the input data are numeric, and which are strings, and which are tuples. This is not always the case. Because ClickHouse is routineley used with datasets containing billions of records there is a default number (100) of rows examined to [infer the schema](../../guides/developer/working-with-json/json-semi-structured.md/#relying-on-schema-inference) in order to avoid parsing billions of rows to infer the schema. The response below may not match what you see, as the dataset is updated several times each year. Looking at the Data Dictionary you can see that CMPLNT_NUM is specified as text, and not numeric. By overriding the default of 100 rows for inference with the setting `SETTINGS input_format_max_rows_to_read_for_schema_inference=2000`
you can get a better idea of the content.
Note: as of version 22.5 the default is now 25,000 rows for inferring the schema, so only change the setting if you are on an older version or if you need more than 25,000 rows to be sampled.
:::
Run this command at your command prompt. You will be using `clickhouse-local` to query the data in the TSV file you downloaded.
```sh
clickhouse-local --input_format_max_rows_to_read_for_schema_inference=2000 \
--query \
"describe file('${HOME}/NYPD_Complaint_Data_Current__Year_To_Date_.tsv', 'TSVWithNames')"
```
Result:
```response
CMPLNT_NUM Nullable(String)
ADDR_PCT_CD Nullable(Float64)
BORO_NM Nullable(String)
CMPLNT_FR_DT Nullable(String)
CMPLNT_FR_TM Nullable(String)
CMPLNT_TO_DT Nullable(String)
CMPLNT_TO_TM Nullable(String)
CRM_ATPT_CPTD_CD Nullable(String)
HADEVELOPT Nullable(String)
HOUSING_PSA Nullable(Float64)
JURISDICTION_CODE Nullable(Float64)
JURIS_DESC Nullable(String)
KY_CD Nullable(Float64)
LAW_CAT_CD Nullable(String)
LOC_OF_OCCUR_DESC Nullable(String)
OFNS_DESC Nullable(String)
PARKS_NM Nullable(String)
PATROL_BORO Nullable(String)
PD_CD Nullable(Float64)
PD_DESC Nullable(String)
PREM_TYP_DESC Nullable(String)
RPT_DT Nullable(String)
STATION_NAME Nullable(String)
SUSP_AGE_GROUP Nullable(String)
SUSP_RACE Nullable(String)
SUSP_SEX Nullable(String)
TRANSIT_DISTRICT Nullable(Float64)
VIC_AGE_GROUP Nullable(String)
VIC_RACE Nullable(String)
VIC_SEX Nullable(String)
X_COORD_CD Nullable(Float64)
Y_COORD_CD Nullable(Float64)
Latitude Nullable(Float64)
Longitude Nullable(Float64)
Lat_Lon Tuple(Nullable(Float64), Nullable(Float64))
New Georeferenced Column Nullable(String)
```
At this point you should check that the columns in the TSV file match the names and types specified in the **Columns in this Dataset** section of the [dataset web page](https://data.cityofnewyork.us/Public-Safety/NYPD-Complaint-Data-Current-Year-To-Date-/5uac-w243). The data types are not very specific, all numeric fields are set to `Nullable(Float64)`, and all other fields are `Nullable(String)`. When you create a ClickHouse table to store the data you can specify more appropriate and performant types.
### Determine the proper schema
In order to figure out what types should be used for the fields it is necessary to know what the data looks like. For example, the field `JURISDICTION_CODE` is a numeric: should it be a `UInt8`, or an `Enum`, or is `Float64` appropriate?
```sql
clickhouse-local --input_format_max_rows_to_read_for_schema_inference=2000 \
--query \
"select JURISDICTION_CODE, count() FROM
file('${HOME}/NYPD_Complaint_Data_Current__Year_To_Date_.tsv', 'TSVWithNames')
GROUP BY JURISDICTION_CODE
ORDER BY JURISDICTION_CODE
FORMAT PrettyCompact"
```
Result:
```response
┌─JURISDICTION_CODE─┬─count()─┐
│ 0 │ 188875 │
│ 1 │ 4799 │
│ 2 │ 13833 │
│ 3 │ 656 │
│ 4 │ 51 │
│ 6 │ 5 │
│ 7 │ 2 │
│ 9 │ 13 │
│ 11 │ 14 │
│ 12 │ 5 │
│ 13 │ 2 │
│ 14 │ 70 │
│ 15 │ 20 │
│ 72 │ 159 │
│ 87 │ 9 │
│ 88 │ 75 │
│ 97 │ 405 │
└───────────────────┴─────────┘
```
The query response shows that the `JURISDICTION_CODE` fits well in a `UInt8`.
Similarly, look at some of the `String` fields and see if they are well suited to being `DateTime` or [`LowCardinality(String)`](../../sql-reference/data-types/lowcardinality.md) fields.
For example, the field `PARKS_NM` is described as "Name of NYC park, playground or greenspace of occurrence, if applicable (state parks are not included)". The names of parks in New York City may be a good candidate for a `LowCardinality(String)`:
```sh
clickhouse-local --input_format_max_rows_to_read_for_schema_inference=2000 \
--query \
"select count(distinct PARKS_NM) FROM
file('${HOME}/NYPD_Complaint_Data_Current__Year_To_Date_.tsv', 'TSVWithNames')
FORMAT PrettyCompact"
```
Result:
```response
┌─uniqExact(PARKS_NM)─┐
│ 319 │
└─────────────────────┘
```
Have a look at some of the park names:
```sql
clickhouse-local --input_format_max_rows_to_read_for_schema_inference=2000 \
--query \
"select distinct PARKS_NM FROM
file('${HOME}/NYPD_Complaint_Data_Current__Year_To_Date_.tsv', 'TSVWithNames')
LIMIT 10
FORMAT PrettyCompact"
```
Result:
```response
┌─PARKS_NM───────────────────┐
│ (null) │
│ ASSER LEVY PARK │
│ JAMES J WALKER PARK │
│ BELT PARKWAY/SHORE PARKWAY │
│ PROSPECT PARK │
│ MONTEFIORE SQUARE │
│ SUTTON PLACE PARK │
│ JOYCE KILMER PARK │
│ ALLEY ATHLETIC PLAYGROUND │
│ ASTORIA PARK │
└────────────────────────────┘
```
The dataset in use at the time of writing has only a few hundred distinct parks and playgrounds in the `PARK_NM` column. This is a small number based on the [LowCardinality](../../sql-reference/data-types/lowcardinality.md#lowcardinality-dscr) recommendation to stay below 10,000 distinct strings in a `LowCardinality(String)` field.
### DateTime fields
Based on the **Columns in this Dataset** section of the [dataset web page](https://data.cityofnewyork.us/Public-Safety/NYPD-Complaint-Data-Current-Year-To-Date-/5uac-w243) there are date and time fields for the start and end of the reported event. Looking at the min and max of the `CMPLNT_FR_DT` and `CMPLT_TO_DT` gives an idea of whether or not the fields are always populated:
```sh title="CMPLNT_FR_DT"
clickhouse-local --input_format_max_rows_to_read_for_schema_inference=2000 \
--query \
"select min(CMPLNT_FR_DT), max(CMPLNT_FR_DT) FROM
file('${HOME}/NYPD_Complaint_Data_Current__Year_To_Date_.tsv', 'TSVWithNames')
FORMAT PrettyCompact"
```
Result:
```response
┌─min(CMPLNT_FR_DT)─┬─max(CMPLNT_FR_DT)─┐
│ 01/01/1973 │ 12/31/2021 │
└───────────────────┴───────────────────┘
```
```sh title="CMPLNT_TO_DT"
clickhouse-local --input_format_max_rows_to_read_for_schema_inference=2000 \
--query \
"select min(CMPLNT_TO_DT), max(CMPLNT_TO_DT) FROM
file('${HOME}/NYPD_Complaint_Data_Current__Year_To_Date_.tsv', 'TSVWithNames')
FORMAT PrettyCompact"
```
Result:
```response
┌─min(CMPLNT_TO_DT)─┬─max(CMPLNT_TO_DT)─┐
│ │ 12/31/2021 │
└───────────────────┴───────────────────┘
```
```sh title="CMPLNT_FR_TM"
clickhouse-local --input_format_max_rows_to_read_for_schema_inference=2000 \
--query \
"select min(CMPLNT_FR_TM), max(CMPLNT_FR_TM) FROM
file('${HOME}/NYPD_Complaint_Data_Current__Year_To_Date_.tsv', 'TSVWithNames')
FORMAT PrettyCompact"
```
Result:
```response
┌─min(CMPLNT_FR_TM)─┬─max(CMPLNT_FR_TM)─┐
│ 00:00:00 │ 23:59:00 │
└───────────────────┴───────────────────┘
```
```sh title="CMPLNT_TO_TM"
clickhouse-local --input_format_max_rows_to_read_for_schema_inference=2000 \
--query \
"select min(CMPLNT_TO_TM), max(CMPLNT_TO_TM) FROM
file('${HOME}/NYPD_Complaint_Data_Current__Year_To_Date_.tsv', 'TSVWithNames')
FORMAT PrettyCompact"
```
Result:
```response
┌─min(CMPLNT_TO_TM)─┬─max(CMPLNT_TO_TM)─┐
│ (null) │ 23:59:00 │
└───────────────────┴───────────────────┘
```
## Make a plan
Based on the above investigation:
- `JURISDICTION_CODE` should be cast as `UInt8`.
- `PARKS_NM` should be cast to `LowCardinality(String)`
- `CMPLNT_FR_DT` and `CMPLNT_FR_TM` are always populated (possibly with a default time of `00:00:00`)
- `CMPLNT_TO_DT` and `CMPLNT_TO_TM` may be empty
- Dates and times are stored in separate fields in the source
- Dates are `mm/dd/yyyy` format
- Times are `hh:mm:ss` format
- Dates and times can be concatenated into DateTime types
- There are some dates before January 1st 1970, which means we need a 64 bit DateTime
:::note
There are many more changes to be made to the types, they all can be determined by following the same investigation steps. Look at the number of distinct strings in a field, the min and max of the numerics, and make your decisions. The table schema that is given later in the guide has many low cardinality strings and unsigned integer fields and very few floating point numerics.
:::
## Concatenate the date and time fields
To concatenate the date and time fields `CMPLNT_FR_DT` and `CMPLNT_FR_TM` into a single `String` that can be cast to a `DateTime`, select the two fields joined by the concatenation operator: `CMPLNT_FR_DT || ' ' || CMPLNT_FR_TM`. The `CMPLNT_TO_DT` and `CMPLNT_TO_TM` fields are handled similarly.
```sh
clickhouse-local --input_format_max_rows_to_read_for_schema_inference=2000 \
--query \
"select CMPLNT_FR_DT || ' ' || CMPLNT_FR_TM AS complaint_begin FROM
file('${HOME}/NYPD_Complaint_Data_Current__Year_To_Date_.tsv', 'TSVWithNames')
LIMIT 10
FORMAT PrettyCompact"
```
Result:
```response
┌─complaint_begin─────┐
│ 07/29/2010 00:01:00 │
│ 12/01/2011 12:00:00 │
│ 04/01/2017 15:00:00 │
│ 03/26/2018 17:20:00 │
│ 01/01/2019 00:00:00 │
│ 06/14/2019 00:00:00 │
│ 11/29/2021 20:00:00 │
│ 12/04/2021 00:35:00 │
│ 12/05/2021 12:50:00 │
│ 12/07/2021 20:30:00 │
└─────────────────────┘
```
## Convert the date and time String to a DateTime64 type
Earlier in the guide we discovered that there are dates in the TSV file before January 1st 1970, which means that we need a 64 bit DateTime type for the dates. The dates also need to be converted from `MM/DD/YYYY` to `YYYY/MM/DD` format. Both of these can be done with [`parseDateTime64BestEffort()`](../../sql-reference/functions/type-conversion-functions.md#parsedatetime64besteffort).
```sh
clickhouse-local --input_format_max_rows_to_read_for_schema_inference=2000 \
--query \
"WITH (CMPLNT_FR_DT || ' ' || CMPLNT_FR_TM) AS CMPLNT_START,
(CMPLNT_TO_DT || ' ' || CMPLNT_TO_TM) AS CMPLNT_END
select parseDateTime64BestEffort(CMPLNT_START) AS complaint_begin,
parseDateTime64BestEffortOrNull(CMPLNT_END) AS complaint_end
FROM file('${HOME}/NYPD_Complaint_Data_Current__Year_To_Date_.tsv', 'TSVWithNames')
ORDER BY complaint_begin ASC
LIMIT 25
FORMAT PrettyCompact"
```
Lines 2 and 3 above contain the concatenation from the previous step, and lines 4 and 5 above parse the strings into `DateTime64`. As the complaint end time is not guaranteed to exist `parseDateTime64BestEffortOrNull` is used.
Result:
```response
┌─────────complaint_begin─┬───────────complaint_end─┐
│ 1925-01-01 10:00:00.000 │ 2021-02-12 09:30:00.000 │
│ 1925-01-01 11:37:00.000 │ 2022-01-16 11:49:00.000 │
│ 1925-01-01 15:00:00.000 │ 2021-12-31 00:00:00.000 │
│ 1925-01-01 15:00:00.000 │ 2022-02-02 22:00:00.000 │
│ 1925-01-01 19:00:00.000 │ 2022-04-14 05:00:00.000 │
│ 1955-09-01 19:55:00.000 │ 2022-08-01 00:45:00.000 │
│ 1972-03-17 11:40:00.000 │ 2022-03-17 11:43:00.000 │
│ 1972-05-23 22:00:00.000 │ 2022-05-24 09:00:00.000 │
│ 1972-05-30 23:37:00.000 │ 2022-05-30 23:50:00.000 │
│ 1972-07-04 02:17:00.000 │ ᴺᵁᴸᴸ │
│ 1973-01-01 00:00:00.000 │ ᴺᵁᴸᴸ │
│ 1975-01-01 00:00:00.000 │ ᴺᵁᴸᴸ │
│ 1976-11-05 00:01:00.000 │ 1988-10-05 23:59:00.000 │
│ 1977-01-01 00:00:00.000 │ 1977-01-01 23:59:00.000 │
│ 1977-12-20 00:01:00.000 │ ᴺᵁᴸᴸ │
│ 1981-01-01 00:01:00.000 │ ᴺᵁᴸᴸ │
│ 1981-08-14 00:00:00.000 │ 1987-08-13 23:59:00.000 │
│ 1983-01-07 00:00:00.000 │ 1990-01-06 00:00:00.000 │
│ 1984-01-01 00:01:00.000 │ 1984-12-31 23:59:00.000 │
│ 1985-01-01 12:00:00.000 │ 1987-12-31 15:00:00.000 │
│ 1985-01-11 09:00:00.000 │ 1985-12-31 12:00:00.000 │
│ 1986-03-16 00:05:00.000 │ 2022-03-16 00:45:00.000 │
│ 1987-01-07 00:00:00.000 │ 1987-01-09 00:00:00.000 │
│ 1988-04-03 18:30:00.000 │ 2022-08-03 09:45:00.000 │
│ 1988-07-29 12:00:00.000 │ 1990-07-27 22:00:00.000 │
└─────────────────────────┴─────────────────────────┘
```
:::note
The dates shown as `1925` above are from errors in the data. There are several records in the original data with dates in the years `1019` - `1022` that should be `2019` - `2022`. They are being stored as Jan 1st 1925 as that is the earliest date with a 64 bit DateTime.
:::
## Create a table
The decisions made above on the data types used for the columns are reflected in the table schema
below. We also need to decide on the `ORDER BY` and `PRIMARY KEY` used for the table. At least one
of `ORDER BY` or `PRIMARY KEY` must be specified. Here are some guidelines on deciding on the
columns to includes in `ORDER BY`, and more information is in the *Next Steps* section at the end
of this document.
### Order By and Primary Key clauses
- The `ORDER BY` tuple should include fields that are used in query filters
- To maximize compression on disk the `ORDER BY` tuple should be ordered by ascending cardinality
- If it exists, the `PRIMARY KEY` tuple must be a subset of the `ORDER BY` tuple
- If only `ORDER BY` is specified, then the same tuple will be used as `PRIMARY KEY`
- The primary key index is created using the `PRIMARY KEY` tuple if specified, otherwise the `ORDER BY` tuple
- The `PRIMARY KEY` index is kept in main memory
Looking at the dataset and the questions that might be answered by querying it we might
decide that we would look at the types of crimes reported over time in the five boroughs of
New York City. These fields might be then included in the `ORDER BY`:
| Column | Description (from the data dictionary) |
| ----------- | --------------------------------------------------- |
| OFNS_DESC | Description of offense corresponding with key code |
| RPT_DT | Date event was reported to police |
| BORO_NM | The name of the borough in which the incident occurred |
Querying the TSV file for the cardinality of the three candidate columns:
```bash
clickhouse-local --input_format_max_rows_to_read_for_schema_inference=2000 \
--query \
"select formatReadableQuantity(uniq(OFNS_DESC)) as cardinality_OFNS_DESC,
formatReadableQuantity(uniq(RPT_DT)) as cardinality_RPT_DT,
formatReadableQuantity(uniq(BORO_NM)) as cardinality_BORO_NM
FROM
file('${HOME}/NYPD_Complaint_Data_Current__Year_To_Date_.tsv', 'TSVWithNames')
FORMAT PrettyCompact"
```
Result:
```response
┌─cardinality_OFNS_DESC─┬─cardinality_RPT_DT─┬─cardinality_BORO_NM─┐
│ 60.00 │ 306.00 │ 6.00 │
└───────────────────────┴────────────────────┴─────────────────────┘
```
Ordering by cardinality, the `ORDER BY` becomes:
```
ORDER BY ( BORO_NM, OFNS_DESC, RPT_DT )
```
:::note
The table below will use more easily read column names, the above names will be mapped to
```
ORDER BY ( borough, offense_description, date_reported )
```
:::
Putting together the changes to data types and the `ORDER BY` tuple gives this table structure:
```sql
CREATE TABLE NYPD_Complaint (
complaint_number String,
precinct UInt8,
borough LowCardinality(String),
complaint_begin DateTime64(0,'America/New_York'),
complaint_end DateTime64(0,'America/New_York'),
was_crime_completed String,
housing_authority String,
housing_level_code UInt32,
jurisdiction_code UInt8,
jurisdiction LowCardinality(String),
offense_code UInt8,
offense_level LowCardinality(String),
location_descriptor LowCardinality(String),
offense_description LowCardinality(String),
park_name LowCardinality(String),
patrol_borough LowCardinality(String),
PD_CD UInt16,
PD_DESC String,
location_type LowCardinality(String),
date_reported Date,
transit_station LowCardinality(String),
suspect_age_group LowCardinality(String),
suspect_race LowCardinality(String),
suspect_sex LowCardinality(String),
transit_district UInt8,
victim_age_group LowCardinality(String),
victim_race LowCardinality(String),
victim_sex LowCardinality(String),
NY_x_coordinate UInt32,
NY_y_coordinate UInt32,
Latitude Float64,
Longitude Float64
) ENGINE = MergeTree
ORDER BY ( borough, offense_description, date_reported )
```
### Finding the primary key of a table
The ClickHouse `system` database, specifically `system.table` has all of the information about the table you
just created. This query shows the `ORDER BY` (sorting key), and the `PRIMARY KEY`:
```sql
SELECT
partition_key,
sorting_key,
primary_key,
table
FROM system.tables
WHERE table = 'NYPD_Complaint'
FORMAT Vertical
```
Response
```response
Query id: 6a5b10bf-9333-4090-b36e-c7f08b1d9e01
Row 1:
──────
partition_key:
sorting_key: borough, offense_description, date_reported
primary_key: borough, offense_description, date_reported
table: NYPD_Complaint
1 row in set. Elapsed: 0.001 sec.
```
## Preprocess and Import Data {#preprocess-import-data}
We will use `clickhouse-local` tool for data preprocessing and `clickhouse-client` to upload it.
### `clickhouse-local` arguments used
:::tip
`table='input'` appears in the arguments to clickhouse-local below. clickhouse-local takes the provided input (`cat ${HOME}/NYPD_Complaint_Data_Current__Year_To_Date_.tsv`) and inserts the input into a table. By default the table is named `table`. In this guide the name of the table is set to `input` to make the data flow clearer. The final argument to clickhouse-local is a query that selects from the table (`FROM input`) which is then piped to `clickhouse-client` to populate the table `NYPD_Complaint`.
:::
```sql
cat ${HOME}/NYPD_Complaint_Data_Current__Year_To_Date_.tsv \
| clickhouse-local --table='input' --input-format='TSVWithNames' \
--input_format_max_rows_to_read_for_schema_inference=2000 \
--query "
WITH (CMPLNT_FR_DT || ' ' || CMPLNT_FR_TM) AS CMPLNT_START,
(CMPLNT_TO_DT || ' ' || CMPLNT_TO_TM) AS CMPLNT_END
SELECT
CMPLNT_NUM AS complaint_number,
ADDR_PCT_CD AS precinct,
BORO_NM AS borough,
parseDateTime64BestEffort(CMPLNT_START) AS complaint_begin,
parseDateTime64BestEffortOrNull(CMPLNT_END) AS complaint_end,
CRM_ATPT_CPTD_CD AS was_crime_completed,
HADEVELOPT AS housing_authority_development,
HOUSING_PSA AS housing_level_code,
JURISDICTION_CODE AS jurisdiction_code,
JURIS_DESC AS jurisdiction,
KY_CD AS offense_code,
LAW_CAT_CD AS offense_level,
LOC_OF_OCCUR_DESC AS location_descriptor,
OFNS_DESC AS offense_description,
PARKS_NM AS park_name,
PATROL_BORO AS patrol_borough,
PD_CD,
PD_DESC,
PREM_TYP_DESC AS location_type,
toDate(parseDateTimeBestEffort(RPT_DT)) AS date_reported,
STATION_NAME AS transit_station,
SUSP_AGE_GROUP AS suspect_age_group,
SUSP_RACE AS suspect_race,
SUSP_SEX AS suspect_sex,
TRANSIT_DISTRICT AS transit_district,
VIC_AGE_GROUP AS victim_age_group,
VIC_RACE AS victim_race,
VIC_SEX AS victim_sex,
X_COORD_CD AS NY_x_coordinate,
Y_COORD_CD AS NY_y_coordinate,
Latitude,
Longitude
FROM input" \
| clickhouse-client --query='INSERT INTO NYPD_Complaint FORMAT TSV'
```
## Validate the Data {#validate-data}
:::note
The dataset changes once or more per year, your counts may not match what is in this document.
:::
Query:
```sql
SELECT count()
FROM NYPD_Complaint
```
Result:
```text
┌─count()─┐
│ 208993 │
└─────────┘
1 row in set. Elapsed: 0.001 sec.
```
The size of the dataset in ClickHouse is just 12% of the original TSV file, compare the size of the original TSV file with the size of the table:
Query:
```sql
SELECT formatReadableSize(total_bytes)
FROM system.tables
WHERE name = 'NYPD_Complaint'
```
Result:
```text
┌─formatReadableSize(total_bytes)─┐
│ 8.63 MiB │
└─────────────────────────────────┘
```
## Run Some Queries {#run-queries}
### Query 1. Compare the number of complaints by month
Query:
```sql
SELECT
dateName('month', date_reported) AS month,
count() AS complaints,
bar(complaints, 0, 50000, 80)
FROM NYPD_Complaint
GROUP BY month
ORDER BY complaints DESC
```
Result:
```response
Query id: 7fbd4244-b32a-4acf-b1f3-c3aa198e74d9
┌─month─────┬─complaints─┬─bar(count(), 0, 50000, 80)───────────────────────────────┐
│ March │ 34536 │ ███████████████████████████████████████████████████████▎ │
│ May │ 34250 │ ██████████████████████████████████████████████████████▋ │
│ April │ 32541 │ ████████████████████████████████████████████████████ │
│ January │ 30806 │ █████████████████████████████████████████████████▎ │
│ February │ 28118 │ ████████████████████████████████████████████▊ │
│ November │ 7474 │ ███████████▊ │
│ December │ 7223 │ ███████████▌ │
│ October │ 7070 │ ███████████▎ │
│ September │ 6910 │ ███████████ │
│ August │ 6801 │ ██████████▊ │
│ June │ 6779 │ ██████████▋ │
│ July │ 6485 │ ██████████▍ │
└───────────┴────────────┴──────────────────────────────────────────────────────────┘
12 rows in set. Elapsed: 0.006 sec. Processed 208.99 thousand rows, 417.99 KB (37.48 million rows/s., 74.96 MB/s.)
```
### Query 2. Compare total number of complaints by Borough
Query:
```sql
SELECT
borough,
count() AS complaints,
bar(complaints, 0, 125000, 60)
FROM NYPD_Complaint
GROUP BY borough
ORDER BY complaints DESC
```
Result:
```response
Query id: 8cdcdfd4-908f-4be0-99e3-265722a2ab8d
┌─borough───────┬─complaints─┬─bar(count(), 0, 125000, 60)──┐
│ BROOKLYN │ 57947 │ ███████████████████████████▋ │
│ MANHATTAN │ 53025 │ █████████████████████████▍ │
│ QUEENS │ 44875 │ █████████████████████▌ │
│ BRONX │ 44260 │ █████████████████████▏ │
│ STATEN ISLAND │ 8503 │ ████ │
│ (null) │ 383 │ ▏ │
└───────────────┴────────────┴──────────────────────────────┘
6 rows in set. Elapsed: 0.008 sec. Processed 208.99 thousand rows, 209.43 KB (27.14 million rows/s., 27.20 MB/s.)
```
## Next Steps
[A Practical Introduction to Sparse Primary Indexes in ClickHouse](../../guides/improving-query-performance/sparse-primary-indexes/sparse-primary-indexes-intro.md) discusses the differences in ClickHouse indexing compared to traditional relational databases, how ClickHouse builds and uses a sparse primary index, and indexing best practices.

122
docs/en/getting-started/example-datasets/uk-price-paid.md
View File

@ -6,19 +6,28 @@ title: "UK Property Price Paid"
---
The dataset contains data about prices paid for real-estate property in England and Wales. The data is available since year 1995.
The size of the dataset in uncompressed form is about 4 GiB and it will take about 270 MiB in ClickHouse.
The size of the dataset in uncompressed form is about 4 GiB and it will take about 278 MiB in ClickHouse.
Source: https://www.gov.uk/government/statistical-data-sets/price-paid-data-downloads <br/>
Source: https://www.gov.uk/government/statistical-data-sets/price-paid-data-downloads
Description of the fields: https://www.gov.uk/guidance/about-the-price-paid-data
Contains HM Land Registry data © Crown copyright and database right 2021. This data is licensed under the Open Government Licence v3.0.
## Download the Dataset {#download-dataset}
Run the command:
```bash
wget http://prod.publicdata.landregistry.gov.uk.s3-website-eu-west-1.amazonaws.com/pp-complete.csv
```
Download will take about 2 minutes with good internet connection.
## Create the Table {#create-table}
```sql
CREATE TABLE uk_price_paid
(
uuid UUID,
price UInt32,
date Date,
postcode1 LowCardinality(String),
@ -33,68 +42,65 @@ CREATE TABLE uk_price_paid
town LowCardinality(String),
district LowCardinality(String),
county LowCardinality(String),
category UInt8,
category2 UInt8
) ORDER BY (postcode1, postcode2, addr1, addr2);
category UInt8
) ENGINE = MergeTree ORDER BY (postcode1, postcode2, addr1, addr2);
```
## Preprocess and Import Data {#preprocess-import-data}
In this example, we define the structure of source data from the CSV file and specify a query to preprocess the data with either `clickhouse-client` or the web based Play UI.
We will use `clickhouse-local` tool for data preprocessing and `clickhouse-client` to upload it.
In this example, we define the structure of source data from the CSV file and specify a query to preprocess the data with `clickhouse-local`.
The preprocessing is:
- splitting the postcode to two different columns `postcode1` and `postcode2` that are better for storage and queries;
- splitting the postcode to two different columns `postcode1` and `postcode2` that is better for storage and queries;
- coverting the `time` field to date as it only contains 00:00 time;
- ignoring the [UUid](../../sql-reference/data-types/uuid.md) field because we don't need it for analysis;
- transforming `type` and `duration` to more readable Enum fields with function [transform](../../sql-reference/functions/other-functions.md#transform);
- transforming `is_new` and `category` fields from single-character string (`Y`/`N` and `A`/`B`) to [UInt8](../../sql-reference/data-types/int-uint.md#uint8-uint16-uint32-uint64-uint256-int8-int16-int32-int64-int128-int256) field with 0 and 1.
Preprocessed data is piped directly to `clickhouse-client` to be inserted into ClickHouse table in streaming fashion.
```bash
INSERT INTO uk_price_paid
WITH
splitByChar(' ', postcode) AS p
SELECT
replaceRegexpAll(uuid_string, '{|}','') AS uuid,
toUInt32(price_string) AS price,
parseDateTimeBestEffortUS(time) AS date,
p[1] AS postcode1,
p[2] AS postcode2,
transform(a, ['T', 'S', 'D', 'F', 'O'], ['terraced', 'semi-detached', 'detached', 'flat', 'other']) AS type,
b = 'Y' AS is_new,
transform(c, ['F', 'L', 'U'], ['freehold', 'leasehold', 'unknown']) AS duration,
addr1,
addr2,
street,
locality,
town,
district,
county,
d = 'B' AS category,
e = 'B' AS category2
FROM url(
'http://prod.publicdata.landregistry.gov.uk.s3-website-eu-west-1.amazonaws.com/pp-complete.csv',
'CSV',
'uuid_string String,
price_string String,
time String,
postcode String,
a String,
b String,
c String,
addr1 String,
addr2 String,
street String,
locality String,
town String,
district String,
county String,
d String,
e String'
)
SETTINGS max_http_get_redirects=1;
clickhouse-local --input-format CSV --structure '
uuid String,
price UInt32,
time DateTime,
postcode String,
a String,
b String,
c String,
addr1 String,
addr2 String,
street String,
locality String,
town String,
district String,
county String,
d String,
e String
' --query "
WITH splitByChar(' ', postcode) AS p
SELECT
price,
toDate(time) AS date,
p[1] AS postcode1,
p[2] AS postcode2,
transform(a, ['T', 'S', 'D', 'F', 'O'], ['terraced', 'semi-detached', 'detached', 'flat', 'other']) AS type,
b = 'Y' AS is_new,
transform(c, ['F', 'L', 'U'], ['freehold', 'leasehold', 'unknown']) AS duration,
addr1,
addr2,
street,
locality,
town,
district,
county,
d = 'B' AS category
FROM table" --date_time_input_format best_effort < pp-complete.csv | clickhouse-client --query "INSERT INTO uk_price_paid FORMAT TSV"
```
It will take about 2 minutes depending on where you are in the world, and where your ClickHouse servers are. Almost all of the time is the download time of the CSV file from the UK government server.
It will take about 40 seconds.
## Validate the Data {#validate-data}
@ -106,13 +112,13 @@ SELECT count() FROM uk_price_paid;
Result:
```response
```text
┌──count()─┐
│ 27450499 │
│ 26321785 │
└──────────┘
```
The size of dataset in ClickHouse is just 540 MiB, check it.
The size of dataset in ClickHouse is just 278 MiB, check it.
Query:
@ -124,14 +130,10 @@ Result:
```text
┌─formatReadableSize(total_bytes)─┐
│ 545.04 MiB │
│ 278.80 MiB │
└─────────────────────────────────┘
```
:::note
The above size is for a replicated table, if you are using this dataset with a single instance the size will be half.
:::
## Run Some Queries {#run-queries}
### Query 1. Average Price Per Year {#average-price}
@ -144,7 +146,7 @@ SELECT toYear(date) AS year, round(avg(price)) AS price, bar(price, 0, 1000000,
Result:
```response
```text
┌─year─┬──price─┬─bar(round(avg(price)), 0, 1000000, 80)─┐
│ 1995 │ 67932 │ █████▍ │
│ 1996 │ 71505 │ █████▋ │

6
docs/en/getting-started/install.md
View File

@ -59,7 +59,7 @@ clickhouse-client # or "clickhouse-client --password" if you set up a password.
</details>
You can replace `stable` with `lts` or `testing` to use different [release trains](../faq/operations/production.md) based on your needs.
You can replace `stable` with `lts` to use different [release kinds](../faq/operations/production.md) based on your needs.
You can also download and install packages manually from [here](https://packages.clickhouse.com/deb/pool/stable).
@ -106,7 +106,7 @@ clickhouse-client # or "clickhouse-client --password" if you set up a password.
</details>
If you want to use the most recent version, replace `stable` with `testing` (this is recommended for your testing environments). `prestable` is sometimes also available.
You can replace `stable` with `lts` to use different [release kinds](../faq/operations/production.md) based on your needs.
Then run these commands to install packages:
@ -221,7 +221,7 @@ For non-Linux operating systems and for AArch64 CPU architecture, ClickHouse bui
curl -O 'https://builds.clickhouse.com/master/aarch64/clickhouse' && chmod a+x ./clickhouse
```
Run `sudo ./clickhouse install` to install ClickHouse system-wide (also with needed configuration files, configuring users etc.). Then run `clickhouse start` commands to start the clickhouse-server and `clickhouse-client` to connect to it.
Run `sudo ./clickhouse install` to install ClickHouse system-wide (also with needed configuration files, configuring users etc.). Then run `sudo clickhouse start` commands to start the clickhouse-server and `clickhouse-client` to connect to it.
Use the `clickhouse client` to connect to the server, or `clickhouse local` to process local data.

4
docs/en/interfaces/http.md
View File

@ -175,6 +175,10 @@ You can also choose to use [HTTP compression](https://en.wikipedia.org/wiki/HTTP
- `br`
- `deflate`
- `xz`
- `zstd`
- `lz4`
- `bz2`
- `snappy`
To send a compressed `POST` request, append the request header `Content-Encoding: compression_method`.
In order for ClickHouse to compress the response, enable compression with [enable_http_compression](../operations/settings/settings.md#settings-enable_http_compression) setting and append `Accept-Encoding: compression_method` header to the request. You can configure the data compression level in the [http_zlib_compression_level](../operations/settings/settings.md#settings-http_zlib_compression_level) setting for all compression methods.

3
docs/en/operations/backup.md
View File

@ -2,10 +2,9 @@
slug: /en/operations/backup
sidebar_position: 49
sidebar_label: Data backup and restore
title: Data backup and restore
---
# Data backup and restore
While [replication](../engines/table-engines/mergetree-family/replication.md) provides protection from hardware failures, it does not protect against human errors: accidental deletion of data, deletion of the wrong table or a table on the wrong cluster, and software bugs that result in incorrect data processing or data corruption. In many cases mistakes like these will affect all replicas. ClickHouse has built-in safeguards to prevent some types of mistakes — for example, by default [you can’t just drop tables with a MergeTree-like engine containing more than 50 Gb of data](server-configuration-parameters/settings.md#max-table-size-to-drop). However, these safeguards do not cover all possible cases and can be circumvented.
In order to effectively mitigate possible human errors, you should carefully prepare a strategy for backing up and restoring your data **in advance**.

4
docs/en/operations/server-configuration-parameters/settings.md
View File

@ -1452,7 +1452,7 @@ Port for communicating with clients over MySQL protocol.
**Possible values**
Positive integer.
Positive integer to specify the port number to listen to or empty value to disable.
Example
@ -1466,7 +1466,7 @@ Port for communicating with clients over PostgreSQL protocol.
**Possible values**
Positive integer.
Positive integer to specify the port number to listen to or empty value to disable.
Example

15
docs/en/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-lifetime.md
View File

@ -94,6 +94,21 @@ It is also possible for `Flat`, `Hashed`, `ComplexKeyHashed` dictionaries to onl
- If the source is HTTP then `update_field` will be added as a query parameter with the last update time as the parameter value.
- If the source is Executable then `update_field` will be added as an executable script argument with the last update time as the argument value.
- If the source is ClickHouse, MySQL, PostgreSQL, ODBC there will be an additional part of `WHERE`, where `update_field` is compared as greater or equal with the last update time.
- Per default, this `WHERE`-condition is checked at the highest level of the SQL-Query. Alternatively, the condition can be checked in any other `WHERE`-clause within the query using the `{condition}`-keyword. Example:
```sql
...
SOURCE(CLICKHOUSE(...
update_field 'added_time'
QUERY '
SELECT my_arr.1 AS x, my_arr.2 AS y, creation_time
FROM (
SELECT arrayZip(x_arr, y_arr) AS my_arr, creation_time
FROM dictionary_source
WHERE {condition}
)'
))
...
```
If `update_field` option is set, additional option `update_lag` can be set. Value of `update_lag` option is subtracted from previous update time before request updated data.

24
docs/en/sql-reference/functions/date-time-functions.md
View File

@ -267,7 +267,7 @@ Result:
└────────────────┘
```
:::Attention
:::note
The return type of `toStartOf*`, `toLastDayOfMonth`, `toMonday` functions described below is `Date` or `DateTime`.
Though these functions can take values of the extended types `Date32` and `DateTime64` as an argument, passing them a time outside the normal range (year 1970 to 2149 for `Date` / 2106 for `DateTime`) will produce wrong results.
In case argument is out of normal range:
@ -640,7 +640,8 @@ Result:
## date\_diff
Returns the difference between two dates or dates with time values.
Returns the difference between two dates or dates with time values.
The difference is calculated using relative units, e.g. the difference between `2022-01-01` and `2021-12-29` is 3 days for day unit (see [toRelativeDayNum](#toRelativeDayNum)), 1 month for month unit (see [toRelativeMonthNum](#toRelativeMonthNum)), 1 year for year unit (see [toRelativeYearNum](#toRelativeYearNum)).
**Syntax**
@ -692,6 +693,25 @@ Result:
└────────────────────────────────────────────────────────────────────────────────────────┘
```
Query:
``` sql
SELECT
toDate('2022-01-01') AS e,
toDate('2021-12-29') AS s,
dateDiff('day', s, e) AS day_diff,
dateDiff('month', s, e) AS month__diff,
dateDiff('year', s, e) AS year_diff;
```
Result:
``` text
┌──────────e─┬──────────s─┬─day_diff─┬─month__diff─┬─year_diff─┐
│ 2022-01-01 │ 2021-12-29 │ 3 │ 1 │ 1 │
└────────────┴────────────┴──────────┴─────────────┴───────────┘
```
## date\_sub
Subtracts the time interval or date interval from the provided date or date with time.

114
docs/en/sql-reference/functions/tuple-map-functions.md
View File

@ -430,5 +430,119 @@ Result:
└────────────────────────────┘
```
## mapApply
**Syntax**
```sql
mapApply(func, map)
```
**Parameters**
- `func` - [Lamda function](../../sql-reference/functions/index.md#higher-order-functions---operator-and-lambdaparams-expr-function).
- `map` — [Map](../../sql-reference/data-types/map.md).
**Returned value**
- Returns a map obtained from the original map by application of `func(map1[i], …, mapN[i])` for each element.
**Example**
Query:
```sql
SELECT mapApply((k, v) -> (k, v * 10), _map) AS r
FROM
(
SELECT map('key1', number, 'key2', number * 2) AS _map
FROM numbers(3)
)
```
Result:
```text
┌─r─────────────────────┐
│ {'key1':0,'key2':0} │
│ {'key1':10,'key2':20} │
│ {'key1':20,'key2':40} │
└───────────────────────┘
```
## mapFilter
**Syntax**
```sql
mapFilter(func, map)
```
**Parameters**
- `func` - [Lamda function](../../sql-reference/functions/index.md#higher-order-functions---operator-and-lambdaparams-expr-function).
- `map` — [Map](../../sql-reference/data-types/map.md).
**Returned value**
- Returns a map containing only the elements in `map` for which `func(map1[i], …, mapN[i])` returns something other than 0.
**Example**
Query:
```sql
SELECT mapFilter((k, v) -> ((v % 2) = 0), _map) AS r
FROM
(
SELECT map('key1', number, 'key2', number * 2) AS _map
FROM numbers(3)
)
```
Result:
```text
┌─r───────────────────┐
│ {'key1':0,'key2':0} │
│ {'key2':2} │
│ {'key1':2,'key2':4} │
└─────────────────────┘
```
## mapUpdate
**Syntax**
```sql
mapUpdate(map1, map2)
```
**Parameters**
- `map1` [Map](../../sql-reference/data-types/map.md).
- `map2` [Map](../../sql-reference/data-types/map.md).
**Returned value**
- Returns a map1 with values updated of values for the corresponding keys in map2.
**Example**
Query:
```sql
SELECT mapUpdate(map('key1', 0, 'key3', 0), map('key1', 10, 'key2', 10)) AS map;
```
Result:
```text
┌─map────────────────────────────┐
│ {'key3':0,'key1':10,'key2':10} │
└────────────────────────────────┘
```
[Original article](https://clickhouse.com/docs/en/sql-reference/functions/tuple-map-functions/) <!--hide-->

94
docs/en/sql-reference/functions/uniqtheta-functions.md Normal file
View File

@ -0,0 +1,94 @@
---
slug: /en/sql-reference/functions/uniqtheta-functions
---
# uniqTheta Functions
uniqTheta functions work for two uniqThetaSketch objects to do set operation calculations such as ∪ / ∩ / × (union/intersect/not), it is to return a new uniqThetaSketch object contain the result.
A uniqThetaSketch object is to be constructed by aggregation function uniqTheta with -State.
UniqThetaSketch is a data structure storage of approximate values set.
For more information on RoaringBitmap, see: [Theta Sketch Framework](https://datasketches.apache.org/docs/Theta/ThetaSketchFramework.html).
## uniqThetaUnion
Two uniqThetaSketch objects to do union calculation(set operation ∪), the result is a new uniqThetaSketch.
``` sql
uniqThetaUnion(uniqThetaSketch,uniqThetaSketch)
```
**Arguments**
- `uniqThetaSketch` – uniqThetaSketch object.
**Example**
``` sql
select finalizeAggregation(uniqThetaUnion(a, b)) as a_union_b, finalizeAggregation(a) as a_cardinality, finalizeAggregation(b) as b_cardinality
from
(select arrayReduce('uniqThetaState',[1,2]) as a, arrayReduce('uniqThetaState',[2,3,4]) as b );
```
``` text
┌─a_union_b─┬─a_cardinality─┬─b_cardinality─┐
│ 4 │ 2 │ 3 │
└───────────┴───────────────┴───────────────┘
```
## uniqThetaIntersect
Two uniqThetaSketch objects to do intersect calculation(set operation ∩), the result is a new uniqThetaSketch.
``` sql
uniqThetaIntersect(uniqThetaSketch,uniqThetaSketch)
```
**Arguments**
- `uniqThetaSketch` – uniqThetaSketch object.
**Example**
``` sql
select finalizeAggregation(uniqThetaIntersect(a, b)) as a_intersect_b, finalizeAggregation(a) as a_cardinality, finalizeAggregation(b) as b_cardinality
from
(select arrayReduce('uniqThetaState',[1,2]) as a, arrayReduce('uniqThetaState',[2,3,4]) as b );
```
``` text
┌─a_intersect_b─┬─a_cardinality─┬─b_cardinality─┐
│ 1 │ 2 │ 3 │
└───────────────┴───────────────┴───────────────┘
```
## uniqThetaNot
Two uniqThetaSketch objects to do a_not_b calculation(set operation ×), the result is a new uniqThetaSketch.
``` sql
uniqThetaNot(uniqThetaSketch,uniqThetaSketch)
```
**Arguments**
- `uniqThetaSketch` – uniqThetaSketch object.
**Example**
``` sql
select finalizeAggregation(uniqThetaNot(a, b)) as a_not_b, finalizeAggregation(a) as a_cardinality, finalizeAggregation(b) as b_cardinality
from
(select arrayReduce('uniqThetaState',[2,3,4]) as a, arrayReduce('uniqThetaState',[1,2]) as b );
```
``` text
┌─a_not_b─┬─a_cardinality─┬─b_cardinality─┐
│ 2 │ 3 │ 2 │
└─────────┴───────────────┴───────────────┘
```
**See Also**
- [uniqThetaSketch](../../sql-reference/aggregate-functions/reference/uniqthetasketch.md#agg_function-uniqthetasketch)

2
docs/en/sql-reference/statements/show.md
View File

@ -303,7 +303,7 @@ SHOW USERS
## SHOW ROLES
Returns a list of [roles](../../operations/access-rights.md#role-management). To view another parameters, see system tables [system.roles](../../operations/system-tables/roles.md#system_tables-roles) and [system.role-grants](../../operations/system-tables/role-grants.md#system_tables-role_grants).
Returns a list of [roles](../../operations/access-rights.md#role-management). To view another parameters, see system tables [system.roles](../../operations/system-tables/roles.md#system_tables-roles) and [system.role_grants](../../operations/system-tables/role-grants.md#system_tables-role_grants).
### Syntax

4
docs/ru/sql-reference/functions/date-time-functions.md
View File

@ -267,7 +267,7 @@ SELECT toUnixTimestamp('2017-11-05 08:07:47', 'Asia/Tokyo') AS unix_timestamp;
└────────────────┘
```
:::Attention
:::note
Тип возвращаемого описанными далее функциями `toStartOf*`, `toMonday` значения - `Date` или `DateTime`.
Хотя эти функции могут принимать значения типа `Date32` или `DateTime64` в качестве аргумента, при обработке аргумента вне нормального диапазона значений (`1970` - `2148` для `Date` и `1970-01-01 00:00:00`-`2106-02-07 08:28:15` для `DateTime`) будет получен некорректный результат.
Возвращаемые значения для значений вне нормального диапазона:
@ -277,7 +277,7 @@ SELECT toUnixTimestamp('2017-11-05 08:07:47', 'Asia/Tokyo') AS unix_timestamp;
* `2149-05-31` будет результатом функции `toLastDayOfMonth` при обработке аргумента больше `2149-05-31`.
:::
:::Attention
:::note
Тип возвращаемого описанными далее функциями `toStartOf*`, `toLastDayOfMonth`, `toMonday` значения - `Date` или `DateTime`.
Хотя эти функции могут принимать значения типа `Date32` или `DateTime64` в качестве аргумента, при обработке аргумента вне нормального диапазона значений (`1970` - `2148` для `Date` и `1970-01-01 00:00:00`-`2106-02-07 08:28:15` для `DateTime`) будет получен некорректный результат.
Возвращаемые значения для значений вне нормального диапазона:

2
docs/ru/sql-reference/statements/show.md
View File

@ -305,7 +305,7 @@ SHOW USERS
## SHOW ROLES {#show-roles-statement}
Выводит список [ролей](../../operations/access-rights.md#role-management). Для просмотра параметров ролей, см. системные таблицы [system.roles](../../operations/system-tables/roles.md#system_tables-roles) и [system.role-grants](../../operations/system-tables/role-grants.md#system_tables-role_grants).
Выводит список [ролей](../../operations/access-rights.md#role-management). Для просмотра параметров ролей, см. системные таблицы [system.roles](../../operations/system-tables/roles.md#system_tables-roles) и [system.role_grants](../../operations/system-tables/role-grants.md#system_tables-role_grants).
### Синтаксис {#show-roles-syntax}

341
docs/zh/development/tests.md
View File

@ -1,338 +1,297 @@
---
slug: /zh/development/tests
slug: /en/development/tests
sidebar_position: 70
sidebar_label: Testing
title: ClickHouse Testing
description: Most of ClickHouse features can be tested with functional tests and they are mandatory to use for every change in ClickHouse code that can be tested that way.
---
# ClickHouse 测试 {#clickhouse-testing}
## 功能测试 {#functional-tests}
## Functional Tests
功能测试使用起来最简单方便. 大多数 ClickHouse 特性都可以通过功能测试进行测试, 并且对于可以通过功能测试进行测试的 ClickHouse 代码的每一个更改, 都必须使用这些特性
Functional tests are the most simple and convenient to use. Most of ClickHouse features can be tested with functional tests and they are mandatory to use for every change in ClickHouse code that can be tested that way.
每个功能测试都会向正在运行的 ClickHouse 服务器发送一个或多个查询, 并将结果与参考进行比较.
Each functional test sends one or multiple queries to the running ClickHouse server and compares the result with reference.
测试位于 `查询` 目录中. 有两个子目录: `无状态` 和 `有状态`. 无状态测试在没有任何预加载测试数据的情况下运行查询 - 它们通常在测试本身内即时创建小型合成数据集. 状态测试需要来自 Yandex.Metrica 的预加载测试数据, 它对公众开放.
Tests are located in `queries` directory. There are two subdirectories: `stateless` and `stateful`. Stateless tests run queries without any preloaded test data - they often create small synthetic datasets on the fly, within the test itself. Stateful tests require preloaded test data from ClickHouse and it is available to general public.
每个测试可以是两种类型之一: `.sql` 和 `.sh`. `.sql` 测试是简单的 SQL 脚本, 它通过管道传输到 `clickhouse-client --multiquery --testmode`. `.sh` 测试是一个自己运行的脚本. SQL 测试通常比 `.sh` 测试更可取. 仅当您必须测试某些无法从纯 SQL 中执行的功能时才应使用 `.sh` 测试, 例如将一些输入数据传送到 `clickhouse-client` 或测试 `clickhouse-local`.
Each test can be one of two types: `.sql` and `.sh`. `.sql` test is the simple SQL script that is piped to `clickhouse-client --multiquery`. `.sh` test is a script that is run by itself. SQL tests are generally preferable to `.sh` tests. You should use `.sh` tests only when you have to test some feature that cannot be exercised from pure SQL, such as piping some input data into `clickhouse-client` or testing `clickhouse-local`.
### 在本地运行测试 {#functional-test-locally}
### Running a Test Locally {#functional-test-locally}
在本地启动ClickHouse服务器, 监听默认端口(9000). 例如, 要运行测试 `01428_hash_set_nan_key`, 请切换到存储库文件夹并运行以下命令:
Start the ClickHouse server locally, listening on the default port (9000). To
run, for example, the test `01428_hash_set_nan_key`, change to the repository
folder and run the following command:
```
PATH=$PATH:<path to clickhouse-client> tests/clickhouse-test 01428_hash_set_nan_key
```
有关更多选项, 请参阅`tests/clickhouse-test --help`. 您可以简单地运行所有测试或运行由测试名称中的子字符串过滤的测试子集：`./clickhouse-test substring`. 还有并行或随机顺序运行测试的选项.
For more options, see `tests/clickhouse-test --help`. You can simply run all tests or run subset of tests filtered by substring in test name: `./clickhouse-test substring`. There are also options to run tests in parallel or in randomized order.
### 添加新测试 {#adding-new-test}
### Adding a New Test
添加新的测试, 在 `queries/0_stateless` 目录下创建 `.sql` 或 `.sh` 文件, 手动检查, 然后通过以下方式生成`.reference`文件：`clickhouse-client -n --testmode < 00000_test.sql > 00000_test.reference` 或 `./00000_test.sh > ./00000_test.reference`.
To add new test, create a `.sql` or `.sh` file in `queries/0_stateless` directory, check it manually and then generate `.reference` file in the following way: `clickhouse-client --multiquery < 00000_test.sql > 00000_test.reference` or `./00000_test.sh > ./00000_test.reference`.
测试应仅使用(创建、删除等)`test` 数据库中假定已预先创建的表; 测试也可以使用临时表.
Tests should use (create, drop, etc) only tables in `test` database that is assumed to be created beforehand; also tests can use temporary tables.
### 选择测试名称 {#choosing-test-name}
### Choosing the Test Name
测试名称以五位数前缀开头, 后跟描述性名称, 例如 `00422_hash_function_constexpr.sql`. 要选择前缀, 请找到目录中已存在的最大前缀, 并将其加一. 在此期间, 可能会添加一些具有相同数字前缀的其他测试, 但这没关系并且不会导致任何问题, 您以后不必更改它.
The name of the test starts with a five-digit prefix followed by a descriptive name, such as `00422_hash_function_constexpr.sql`. To choose the prefix, find the largest prefix already present in the directory, and increment it by one. In the meantime, some other tests might be added with the same numeric prefix, but this is OK and does not lead to any problems, you don't have to change it later.
一些测试的名称中标有 `zookeeper`、`shard` 或 `long` . `zookeeper` 用于使用 ZooKeeper 的测试. `shard` 用于需要服务器监听 `127.0.0.*` 的测试; `distributed` 或 `global` 具有相同的含义. `long` 用于运行时间稍长于一秒的测试. Yo你可以分别使用 `--no-zookeeper`、`--no-shard` 和 `--no-long` 选项禁用这些测试组. 如果需要 ZooKeeper 或分布式查询，请确保为您的测试名称添加适当的前缀.
Some tests are marked with `zookeeper`, `shard` or `long` in their names. `zookeeper` is for tests that are using ZooKeeper. `shard` is for tests that requires server to listen `127.0.0.*`; `distributed` or `global` have the same meaning. `long` is for tests that run slightly longer that one second. You can disable these groups of tests using `--no-zookeeper`, `--no-shard` and `--no-long` options, respectively. Make sure to add a proper prefix to your test name if it needs ZooKeeper or distributed queries.
### 检查必须发生的错误 {#checking-error-must-occur}
### Checking for an Error that Must Occur
有时您想测试是否因不正确的查询而发生服务器错误. 我们支持在 SQL 测试中对此进行特殊注释, 形式如下:
Sometimes you want to test that a server error occurs for an incorrect query. We support special annotations for this in SQL tests, in the following form:
```
select x; -- { serverError 49 }
```
此测试确保服务器返回关于未知列“x”的错误代码为 49. 如果没有错误, 或者错误不同, 则测试失败. 如果您想确保错误发生在客户端, 请改用 `clientError` 注释.
This test ensures that the server returns an error with code 49 about unknown column `x`. If there is no error, or the error is different, the test will fail. If you want to ensure that an error occurs on the client side, use `clientError` annotation instead.
不要检查错误消息的特定措辞, 它将来可能会发生变化, 并且测试将不必要地中断. 只检查错误代码. 如果现有的错误代码不足以满足您的需求, 请考虑添加一个新的.
Do not check for a particular wording of error message, it may change in the future, and the test will needlessly break. Check only the error code. If the existing error code is not precise enough for your needs, consider adding a new one.
### 测试分布式查询 {#testing-distributed-query}
### Testing a Distributed Query
如果你想在功能测试中使用分布式查询, 你可以使用 `127.0.0.{1..2}` 的地址, 以便服务器查询自己; 或者您可以在服务器配置文件中使用预定义的测试集群, 例如`test_shard_localhost`. 请记住在测试名称中添加 `shard` 或 `distributed` 字样, 以便它以正确的配置在 CI 中运行, 其中服务器配置为支持分布式查询.
If you want to use distributed queries in functional tests, you can leverage `remote` table function with `127.0.0.{1..2}` addresses for the server to query itself; or you can use predefined test clusters in server configuration file like `test_shard_localhost`. Remember to add the words `shard` or `distributed` to the test name, so that it is run in CI in correct configurations, where the server is configured to support distributed queries.
## 已知错误 {#known-bugs}
## Known Bugs {#known-bugs}
如果我们知道一些可以通过功能测试轻松重现的错误, 我们将准备好的功能测试放在 `tests/queries/bugs` 目录中. 修复错误后, 这些测试将移至 `tests/queries/0_stateless` .
If we know some bugs that can be easily reproduced by functional tests, we place prepared functional tests in `tests/queries/bugs` directory. These tests will be moved to `tests/queries/0_stateless` when bugs are fixed.
## 集成测试 {#integration-tests}
## Integration Tests {#integration-tests}
集成测试允许在集群配置中测试 ClickHouse 以及 ClickHouse 与其他服务器(如 MySQL、Postgres、MongoDB)的交互. 它们可以用来模拟网络分裂、丢包等情况. 这些测试在Docker下运行, 并使用各种软件创建多个容器.
Integration tests allow testing ClickHouse in clustered configuration and ClickHouse interaction with other servers like MySQL, Postgres, MongoDB. They are useful to emulate network splits, packet drops, etc. These tests are run under Docker and create multiple containers with various software.
有关如何运行这些测试, 请参阅 `tests/integration/README.md` .
See `tests/integration/README.md` on how to run these tests.
注意, ClickHouse与第三方驱动程序的集成没有经过测试. 另外, 我们目前还没有JDBC和ODBC驱动程序的集成测试.
Note that integration of ClickHouse with third-party drivers is not tested. Also, we currently do not have integration tests with our JDBC and ODBC drivers.
## 单元测试 {#unit-tests}
## Unit Tests {#unit-tests}
当您想测试的不是 ClickHouse 整体, 而是单个独立库或类时，单元测试很有用. 您可以使用 `ENABLE_TESTS` CMake 选项启用或禁用测试构建. 单元测试(和其他测试程序)位于代码中的 `tests` 子目录中. 要运行单元测试, 请键入 `ninja test` 。有些测试使用 `gtest` , 但有些程序在测试失败时会返回非零退出码.
Unit tests are useful when you want to test not the ClickHouse as a whole, but a single isolated library or class. You can enable or disable build of tests with `ENABLE_TESTS` CMake option. Unit tests (and other test programs) are located in `tests` subdirectories across the code. To run unit tests, type `ninja test`. Some tests use `gtest`, but some are just programs that return non-zero exit code on test failure.
如果代码已经被功能测试覆盖了, 就没有必要进行单元测试(而且功能测试通常更易于使用).
It’s not necessary to have unit tests if the code is already covered by functional tests (and functional tests are usually much more simple to use).
例如, 您可以通过直接调用可执行文件来运行单独的 gtest 检查:
You can run individual gtest checks by calling the executable directly, for example:
```bash
$ ./src/unit_tests_dbms --gtest_filter=LocalAddress*
```
## 性能测试 {#performance-tests}
## Performance Tests {#performance-tests}
性能测试允许测量和比较 ClickHouse 的某些孤立部分在合成查询上的性能. 测试位于 `tests/performance`. 每个测试都由带有测试用例描述的 `.xml` 文件表示. 测试使用 `docker/tests/performance-comparison` 工具运行. 请参阅自述文件以进行调用.
Performance tests allow to measure and compare performance of some isolated part of ClickHouse on synthetic queries. Performance tests are located at `tests/performance/`. Each test is represented by an `.xml` file with a description of the test case. Tests are run with `docker/test/performance-comparison` tool . See the readme file for invocation.
每个测试在循环中运行一个或多个查询(可能带有参数组合). 一些测试可以包含预加载测试数据集的先决条件.
Each test run one or multiple queries (possibly with combinations of parameters) in a loop.
如果您希望在某些场景中提高ClickHouse的性能，并且如果可以在简单的查询中观察到改进，那么强烈建议编写性能测试。在测试期间使用 `perf top` 或其他perf工具总是有意义的.
If you want to improve performance of ClickHouse in some scenario, and if improvements can be observed on simple queries, it is highly recommended to write a performance test. Also, it is recommended to write performance tests when you add or modify SQL functions which are relatively isolated and not too obscure. It always makes sense to use `perf top` or other `perf` tools during your tests.
## 测试工具和脚本 {#test-tools-and-scripts}
## Test Tools and Scripts {#test-tools-and-scripts}
`tests` 目录中的一些程序不是准备好的测试，而是测试工具. 例如, 对于 `Lexer`, 有一个工具 `src/Parsers/tests/lexer` , 它只是对标准输入进行标记化并将着色结果写入标准输出. 您可以将这些类型的工具用作代码示例以及用于探索和手动测试.
Some programs in `tests` directory are not prepared tests, but are test tools. For example, for `Lexer` there is a tool `src/Parsers/tests/lexer` that just do tokenization of stdin and writes colorized result to stdout. You can use these kind of tools as a code examples and for exploration and manual testing.
## 其他测试 {#miscellaneous-tests}
## Miscellaneous Tests {#miscellaneous-tests}
在 `tests/external_models` 中有机器学习模型的测试. 这些测试不会更新, 必须转移到集成测试.
There are tests for machine learned models in `tests/external_models`. These tests are not updated and must be transferred to integration tests.
仲裁插入有单独的测试. 该测试在不同的服务器上运行 ClickHouse 集群并模拟各种故障情况：网络分裂、丢包(ClickHouse 节点之间、ClickHouse 和 ZooKeeper 之间、ClickHouse 服务器和客户端之间等)、`kill -9`、`kill -STOP` 和 `kill -CONT` , 比如 [Jepsen](https://aphyr.com/tags/Jepsen). 然后测试检查所有已确认的插入是否已写入并且所有被拒绝的插入均未写入.
There is separate test for quorum inserts. This test run ClickHouse cluster on separate servers and emulate various failure cases: network split, packet drop (between ClickHouse nodes, between ClickHouse and ZooKeeper, between ClickHouse server and client, etc.), `kill -9`, `kill -STOP` and `kill -CONT` , like [Jepsen](https://aphyr.com/tags/Jepsen). Then the test checks that all acknowledged inserts was written and all rejected inserts was not.
在 ClickHouse 开源之前, Quorum 测试是由单独的团队编写的. 这个团队不再与ClickHouse合作. 测试碰巧是用Java编写的. 由于这些原因, 必须重写仲裁测试并将其转移到集成测试.
Quorum test was written by separate team before ClickHouse was open-sourced. This team no longer work with ClickHouse. Test was accidentally written in Java. For these reasons, quorum test must be rewritten and moved to integration tests.
## 手动测试 {#manual-testing}
## Manual Testing {#manual-testing}
当您开发一个新特性时, 手动测试它也是合理的. 您可以按照以下步骤进行操作:
When you develop a new feature, it is reasonable to also test it manually. You can do it with the following steps:
构建 ClickHouse. 从终端运行 ClickHouse：将目录更改为 `programs/clickhouse-server` 并使用 `./clickhouse-server` 运行它. 默认情况下, 它将使用当前目录中的配置(`config.xml`、`users.xml` 和`config.d` 和`users.d` 目录中的文件). 要连接到 ClickHouse 服务器, 请运行 `programs/clickhouse-client/clickhouse-client` .
Build ClickHouse. Run ClickHouse from the terminal: change directory to `programs/clickhouse-server` and run it with `./clickhouse-server`. It will use configuration (`config.xml`, `users.xml` and files within `config.d` and `users.d` directories) from the current directory by default. To connect to ClickHouse server, run `programs/clickhouse-client/clickhouse-client`.
请注意, 所有 clickhouse 工具(服务器、客户端等)都只是指向名为 `clickhouse` 的单个二进制文件的符号链接. 你可以在 `programs/clickhouse` 找到这个二进制文件. 所有工具也可以作为 `clickhouse tool` 而不是 `clickhouse-tool` 调用.
Note that all clickhouse tools (server, client, etc) are just symlinks to a single binary named `clickhouse`. You can find this binary at `programs/clickhouse`. All tools can also be invoked as `clickhouse tool` instead of `clickhouse-tool`.
或者, 您可以安装 ClickHouse 包: 从 Yandex 存储库稳定发布, 或者您可以在 ClickHouse 源根目录中使用 `./release` 为自己构建包. 然后使用 `sudo service clickhouse-server start` 启动服务器(或停止以停止服务器). 在 `/etc/clickhouse-server/clickhouse-server.log` 中查找日志.
Alternatively you can install ClickHouse package: either stable release from ClickHouse repository or you can build package for yourself with `./release` in ClickHouse sources root. Then start the server with `sudo clickhouse start` (or stop to stop the server). Look for logs at `/etc/clickhouse-server/clickhouse-server.log`.
当您的系统上已经安装了 ClickHouse 时，您可以构建一个新的 `clickhouse` 二进制文件并替换现有的二进制文件:
When ClickHouse is already installed on your system, you can build a new `clickhouse` binary and replace the existing binary:
``` bash
$ sudo service clickhouse-server stop
$ sudo clickhouse stop
$ sudo cp ./clickhouse /usr/bin/
$ sudo service clickhouse-server start
$ sudo clickhouse start
```
您也可以停止系统 clickhouse-server 并使用相同的配置运行您自己的服务器, 但登录到终端:
Also you can stop system clickhouse-server and run your own with the same configuration but with logging to terminal:
``` bash
$ sudo service clickhouse-server stop
$ sudo clickhouse stop
$ sudo -u clickhouse /usr/bin/clickhouse server --config-file /etc/clickhouse-server/config.xml
```
使用 gdb 的示例:
Example with gdb:
``` bash
$ sudo -u clickhouse gdb --args /usr/bin/clickhouse server --config-file /etc/clickhouse-server/config.xml
```
如果系统 clickhouse-server 已经在运行并且你不想停止它, 你可以在你的 `config.xml` 中更改端口号(或在 `config.d` 目录中的文件中覆盖它们), 提供适当的数据路径, 并运行它.
If the system clickhouse-server is already running and you do not want to stop it, you can change port numbers in your `config.xml` (or override them in a file in `config.d` directory), provide appropriate data path, and run it.
`clickhouse` 二进制文件几乎没有依赖关系, 可以在广泛的 Linux 发行版中使用. 要在服务器上快速而肮脏地测试您的更改, 您可以简单地将新构建的 `clickhouse` 二进制文件 `scp` 到您的服务器, 然后按照上面的示例运行它.
`clickhouse` binary has almost no dependencies and works across wide range of Linux distributions. To quick and dirty test your changes on a server, you can simply `scp` your fresh built `clickhouse` binary to your server and then run it as in examples above.
## 测试环境 {#testing-environment}
## Build Tests {#build-tests}
在发布稳定版之前, 我们将其部署在测试环境中.测试环境是一个集群，处理 [Yandex.Metrica](https://metrica.yandex.com/) 数据的 1/39 部分. 我们与 Yandex.Metrica 团队共享我们的测试环境. ClickHouse无需在现有数据上停机即可升级. 我们首先看到的是, 数据被成功地处理了, 没有滞后于实时, 复制继续工作, Yandex.Metrica 团队没有发现任何问题. 第一次检查可以通过以下方式进行:
Build tests allow to check that build is not broken on various alternative configurations and on some foreign systems. These tests are automated as well.
``` sql
SELECT hostName() AS h, any(version()), any(uptime()), max(UTCEventTime), count() FROM remote('example01-01-{1..3}t', merge, hits) WHERE EventDate >= today() - 2 GROUP BY h ORDER BY h;
```
Examples:
- cross-compile for Darwin x86_64 (Mac OS X)
- cross-compile for FreeBSD x86_64
- cross-compile for Linux AArch64
- build on Ubuntu with libraries from system packages (discouraged)
- build with shared linking of libraries (discouraged)
在某些情况下, 我们还会部署到 Yandex 中我们朋友团队的测试环境：Market、Cloud 等. 此外, 我们还有一些用于开发目的的硬件服务器.
For example, build with system packages is bad practice, because we cannot guarantee what exact version of packages a system will have. But this is really needed by Debian maintainers. For this reason we at least have to support this variant of build. Another example: shared linking is a common source of trouble, but it is needed for some enthusiasts.
## 负载测试 {#load-testing}
Though we cannot run all tests on all variant of builds, we want to check at least that various build variants are not broken. For this purpose we use build tests.
部署到测试环境后, 我们使用来自生产集群的查询运行负载测试. 这是手动完成的.
We also test that there are no translation units that are too long to compile or require too much RAM.
确保您在生产集群上启用了 `query_log`.
We also test that there are no too large stack frames.
收集一天或更长时间的查询日志:
## Testing for Protocol Compatibility {#testing-for-protocol-compatibility}
``` bash
$ clickhouse-client --query="SELECT DISTINCT query FROM system.query_log WHERE event_date = today() AND query LIKE '%ym:%' AND query NOT LIKE '%system.query_log%' AND type = 2 AND is_initial_query" > queries.tsv
```
When we extend ClickHouse network protocol, we test manually that old clickhouse-client works with new clickhouse-server and new clickhouse-client works with old clickhouse-server (simply by running binaries from corresponding packages).
这是一个复杂的例子. `type = 2` 将过滤成功执行的查询. `query LIKE '%ym:%'` 是从 Yandex.Metrica 中选择相关查询. `is_initial_query` 是只选择客户端发起的查询, 而不是 ClickHouse 本身(作为分布式查询处理的一部分).
We also test some cases automatically with integrational tests:
- if data written by old version of ClickHouse can be successfully read by the new version;
- do distributed queries work in a cluster with different ClickHouse versions.
`scp` 将此日志记录到您的测试集群并按如下方式运行它:
## Help from the Compiler {#help-from-the-compiler}
``` bash
$ clickhouse benchmark --concurrency 16 < queries.tsv
```
Main ClickHouse code (that is located in `dbms` directory) is built with `-Wall -Wextra -Werror` and with some additional enabled warnings. Although these options are not enabled for third-party libraries.
(可能你还想指定一个 `--user`)
Clang has even more useful warnings - you can look for them with `-Weverything` and pick something to default build.
然后把它留到晚上或周末, 去休息一下.
For production builds, clang is used, but we also test make gcc builds. For development, clang is usually more convenient to use. You can build on your own machine with debug mode (to save battery of your laptop), but please note that compiler is able to generate more warnings with `-O3` due to better control flow and inter-procedure analysis. When building with clang in debug mode, debug version of `libc++` is used that allows to catch more errors at runtime.
您应该检查 `clickhouse-server` 没有崩溃, 内存占用是有限的, 且性能不会随着时间的推移而降低.
## Sanitizers {#sanitizers}
由于查询和环境的高度可变性, 没有记录和比较精确的查询执行时间.
### Address sanitizer
We run functional, integration, stress and unit tests under ASan on per-commit basis.
## 构建测试 {#build-tests}
### Thread sanitizer
We run functional, integration, stress and unit tests under TSan on per-commit basis.
构建测试允许检查在各种可选配置和一些外部系统上的构建是否被破坏. 这些测试也是自动化的.
### Memory sanitizer
We run functional, integration, stress and unit tests under MSan on per-commit basis.
示例:
- Darwin x86_64 (Mac OS X) 交叉编译
- FreeBSD x86_64 交叉编译
- Linux AArch64 交叉编译
- 使用系统包中的库在 Ubuntu 上构建（不鼓励）
- 使用库的共享链接构建（不鼓励）
例如, 使用系统包构建是不好的做法, 因为我们无法保证系统将拥有哪个确切版本的包. 但这确实是 Debian 维护者所需要的. 出于这个原因, 我们至少必须支持这种构建变体. 另一个例子: 共享链接是一个常见的麻烦来源, 但对于一些爱好者来说是需要的.
虽然我们无法对所有构建变体运行所有测试, 但我们希望至少检查各种构建变体没有被破坏. 为此, 我们使用构建测试.
我们还测试了那些太长而无法编译或需要太多RAM的没有翻译单元.
我们还测试没有太大的堆栈帧.
## 协议兼容性测试 {#testing-for-protocol-compatibility}
当我们扩展 ClickHouse 网络协议时, 我们手动测试旧的 clickhouse-client 与新的 clickhouse-server 一起工作, 而新的 clickhouse-client 与旧的 clickhouse-server 一起工作(只需从相应的包中运行二进制文件).
我们还使用集成测试自动测试一些案例:
- 旧版本ClickHouse写入的数据是否可以被新版本成功读取;
- 在具有不同 ClickHouse 版本的集群中执行分布式查询.
## 编译器的帮助 {#help-from-the-compiler}
主要的 ClickHouse 代码(位于 `dbms` 目录中)是用 `-Wall -Wextra -Werror` 和一些额外的启用警告构建的. 虽然没有为第三方库启用这些选项.
Clang 有更多有用的警告 - 你可以用 `-Weverything` 寻找它们并选择一些东西来默认构建.
对于生产构建, 使用 clang, 但我们也测试 make gcc 构建. 对于开发, clang 通常使用起来更方便. 您可以使用调试模式在自己的机器上构建(以节省笔记本电脑的电池), 但请注意, 由于更好的控制流和过程间分析, 编译器能够使用 `-O3` 生成更多警告. 在调试模式下使用 clang 构建时, 使用调试版本的 `libc++` 允许在运行时捕获更多错误.
## 地址清理器 {#sanitizers}
### 地址清理器
我们在ASan上运行功能测试、集成测试、压力测试和单元测试.
### 线程清理器
我们在TSan下运行功能测试、集成测试、压力测试和单元测试.
### 内存清理器
我们在MSan上运行功能测试、集成测试、压力测试和单元测试.
### 未定义的行为清理器
我们在UBSan下运行功能测试、集成测试、压力测试和单元测试. 某些第三方库的代码未针对 UB 进行清理.
### Undefined behaviour sanitizer
We run functional, integration, stress and unit tests under UBSan on per-commit basis. The code of some third-party libraries is not sanitized for UB.
### Valgrind (Memcheck)
我们曾经在 Valgrind 下通宵运行功能测试, 但不再这样做了. 这需要几个小时. 目前在`re2`库中有一个已知的误报, 见[这篇文章](https://research.swtch.com/sparse).
We used to run functional tests under Valgrind overnight, but don't do it anymore. It takes multiple hours. Currently there is one known false positive in `re2` library, see [this article](https://research.swtch.com/sparse).
## 模糊测试 {#fuzzing}
## Fuzzing {#fuzzing}
ClickHouse 模糊测试是使用 [libFuzzer](https://llvm.org/docs/LibFuzzer.html) 和随机 SQL 查询实现的. 所有模糊测试都应使用sanitizers(地址和未定义)进行.
ClickHouse fuzzing is implemented both using [libFuzzer](https://llvm.org/docs/LibFuzzer.html) and random SQL queries.
All the fuzz testing should be performed with sanitizers (Address and Undefined).
LibFuzzer 用于库代码的隔离模糊测试. Fuzzer 作为测试代码的一部分实现, 并具有 `_fuzzer` 名称后缀.
Fuzzer 示例可以在 `src/Parsers/tests/lexer_fuzzer.cpp` 中找到. LibFuzzer 特定的配置、字典和语料库存储在 `tests/fuzz` 中.
我们鼓励您为处理用户输入的每个功能编写模糊测试.
LibFuzzer is used for isolated fuzz testing of library code. Fuzzers are implemented as part of test code and have “_fuzzer” name postfixes.
Fuzzer example can be found at `src/Parsers/fuzzers/lexer_fuzzer.cpp`. LibFuzzer-specific configs, dictionaries and corpus are stored at `tests/fuzz`.
We encourage you to write fuzz tests for every functionality that handles user input.
默认情况下不构建模糊器. 要构建模糊器, 应设置` -DENABLE_FUZZING=1` 和 `-DENABLE_TESTS=1` 选项.
我们建议在构建模糊器时禁用 Jemalloc. 用于将 ClickHouse fuzzing 集成到 Google OSS-Fuzz 的配置可以在 `docker/fuzz` 中找到.
Fuzzers are not built by default. To build fuzzers both `-DENABLE_FUZZING=1` and `-DENABLE_TESTS=1` options should be set.
We recommend to disable Jemalloc while building fuzzers. Configuration used to integrate ClickHouse fuzzing to
Google OSS-Fuzz can be found at `docker/fuzz`.
我们还使用简单的模糊测试来生成随机SQL查询, 并检查服务器在执行这些查询时是否会死亡.
你可以在 `00746_sql_fuzzy.pl` 中找到它. 这个测试应该连续运行(通宵或更长时间).
We also use simple fuzz test to generate random SQL queries and to check that the server does not die executing them.
You can find it in `00746_sql_fuzzy.pl`. This test should be run continuously (overnight and longer).
我们还使用复杂的基于 AST 的查询模糊器, 它能够找到大量的极端情况. 它在查询 AST 中进行随机排列和替换. 它会记住先前测试中的 AST 节点, 以使用它们对后续测试进行模糊测试, 同时以随机顺序处理它们. 您可以在 [这篇博客文章](https://clickhouse.com/blog/en/2021/fuzzing-clickhouse/) 中了解有关此模糊器的更多信息.
We also use sophisticated AST-based query fuzzer that is able to find huge amount of corner cases. It does random permutations and substitutions in queries AST. It remembers AST nodes from previous tests to use them for fuzzing of subsequent tests while processing them in random order. You can learn more about this fuzzer in [this blog article](https://clickhouse.com/blog/en/2021/fuzzing-clickhouse/).
## 压力测试 {#stress-test}
## Stress test
压力测试是另一种模糊测试. 它使用单个服务器以随机顺序并行运行所有功能测试. 不检查测试结果.
Stress tests are another case of fuzzing. It runs all functional tests in parallel in random order with a single server. Results of the tests are not checked.
经检查:
- 服务器不会崩溃，不会触发调试或清理程序陷阱;
- 没有死锁;
- 数据库结构一致;
- 服务器可以在测试后成功停止并重新启动，没有异常;
It is checked that:
- server does not crash, no debug or sanitizer traps are triggered;
- there are no deadlocks;
- the database structure is consistent;
- server can successfully stop after the test and start again without exceptions.
有五种变体 (Debug, ASan, TSan, MSan, UBSan).
There are five variants (Debug, ASan, TSan, MSan, UBSan).
## 线程模糊器 {#thread-fuzzer}
## Thread Fuzzer
Thread Fuzzer(请不要与 Thread Sanitizer 混淆)是另一种允许随机化线程执行顺序的模糊测试. 它有助于找到更多特殊情况.
Thread Fuzzer (please don't mix up with Thread Sanitizer) is another kind of fuzzing that allows to randomize thread order of execution. It helps to find even more special cases.
## 安全审计 {#security-audit}
## Security Audit
Yandex安全团队的人员从安全的角度对ClickHouse的功能做了一些基本的概述.
Our Security Team did some basic overview of ClickHouse capabilities from the security standpoint.
## 静态分析仪 {#static-analyzers}
## Static Analyzers {#static-analyzers}
我们在每次提交的基础上运行 `clang-tidy`. `clang-static-analyzer` 检查也被启用. `clang-tidy` 也用于一些样式检查.
We run `clang-tidy` on per-commit basis. `clang-static-analyzer` checks are also enabled. `clang-tidy` is also used for some style checks.
我们已经评估了 `clang-tidy`、`Coverity`、`cppcheck`、`PVS-Studio`、`tscancode`、`CodeQL`. 您将在 `tests/instructions/` 目录中找到使用说明. 你也可以阅读[俄文文章](https://habr.com/company/yandex/blog/342018/).
We have evaluated `clang-tidy`, `Coverity`, `cppcheck`, `PVS-Studio`, `tscancode`, `CodeQL`. You will find instructions for usage in `tests/instructions/` directory.
如果你使用 `CLion` 作为 IDE, 你可以利用一些开箱即用的 `clang-tidy` 检查
If you use `CLion` as an IDE, you can leverage some `clang-tidy` checks out of the box.
我们还使用 `shellcheck` 对shell脚本进行静态分析.
We also use `shellcheck` for static analysis of shell scripts.
## 硬化 {#hardening}
## Hardening {#hardening}
在调试版本中, 我们使用自定义分配器执行用户级分配的 ASLR.
In debug build we are using custom allocator that does ASLR of user-level allocations.
我们还手动保护在分配后预期为只读的内存区域.
We also manually protect memory regions that are expected to be readonly after allocation.
在调试构建中, 我们还需要对libc进行自定义, 以确保不会调用 "有害的" (过时的、不安全的、非线程安全的)函数.
In debug build we also involve a customization of libc that ensures that no "harmful" (obsolete, insecure, not thread-safe) functions are called.
Debug 断言被广泛使用.
Debug assertions are used extensively.
在调试版本中，如果抛出带有 "逻辑错误" 代码(暗示错误)的异常, 则程序会过早终止. 它允许在发布版本中使用异常, 但在调试版本中使其成为断言.
In debug build, if exception with "logical error" code (implies a bug) is being thrown, the program is terminated prematurely. It allows to use exceptions in release build but make it an assertion in debug build.
jemalloc 的调试版本用于调试版本.
libc++ 的调试版本用于调试版本.
Debug version of jemalloc is used for debug builds.
Debug version of libc++ is used for debug builds.
## 运行时完整性检查
## Runtime Integrity Checks
对存储在磁盘上的数据是校验和. MergeTree 表中的数据同时以三种方式进行校验和*(压缩数据块、未压缩数据块、跨块的总校验和). 客户端和服务器之间或服务器之间通过网络传输的数据也会进行校验和. 复制确保副本上的数据位相同.
Data stored on disk is checksummed. Data in MergeTree tables is checksummed in three ways simultaneously* (compressed data blocks, uncompressed data blocks, the total checksum across blocks). Data transferred over network between client and server or between servers is also checksummed. Replication ensures bit-identical data on replicas.
需要防止硬件故障(存储介质上的位腐烂、服务器上 RAM 中的位翻转、网络控制器 RAM 中的位翻转、网络交换机 RAM 中的位翻转、客户端 RAM 中的位翻转、线路上的位翻转). 请注意，比特位操作很常见, 即使对于 ECC RAM 和 TCP 校验和(如果您每天设法运行数千台处理 PB 数据的服务器, 也可能发生比特位操作. [观看视频(俄语)](https://www.youtube.com/watch?v=ooBAQIe0KlQ).
It is required to protect from faulty hardware (bit rot on storage media, bit flips in RAM on server, bit flips in RAM of network controller, bit flips in RAM of network switch, bit flips in RAM of client, bit flips on the wire). Note that bit flips are common and likely to occur even for ECC RAM and in presence of TCP checksums (if you manage to run thousands of servers processing petabytes of data each day). [See the video (russian)](https://www.youtube.com/watch?v=ooBAQIe0KlQ).
ClickHouse 提供诊断功能, 可帮助运维工程师找到故障硬件.
ClickHouse provides diagnostics that will help ops engineers to find faulty hardware.
\* 它并不慢.
\* and it is not slow.
## 代码风格 {#code-style}
## Code Style {#code-style}
[此处](style.md)描述了代码样式规则.
Code style rules are described [here](style.md).
要检查一些常见的样式违规，您可以使用 `utils/check-style` 脚本.
To check for some common style violations, you can use `utils/check-style` script.
要强制使用正确的代码样式, 您可以使用 `clang-format`. 文件 `.clang-format` 位于源根目录. 它大多与我们的实际代码风格相对应. 但是不建议将 `clang-format` 应用于现有文件, 因为它会使格式变得更糟. 您可以使用可以在 clang 源代码库中找到的 `clang-format-diff` 工具.
To force proper style of your code, you can use `clang-format`. File `.clang-format` is located at the sources root. It mostly corresponding with our actual code style. But it’s not recommended to apply `clang-format` to existing files because it makes formatting worse. You can use `clang-format-diff` tool that you can find in clang source repository.
或者, 您可以尝试使用 `uncrustify` 工具来重新格式化您的代码. 配置位于源根目录中的 `uncrustify.cfg` 中. 它比 `clang-format` 测试更少.
Alternatively you can try `uncrustify` tool to reformat your code. Configuration is in `uncrustify.cfg` in the sources root. It is less tested than `clang-format`.
`CLion` 有自己的代码格式化程序, 必须根据我们的代码风格进行调整.
`CLion` has its own code formatter that has to be tuned for our code style.
我们还使用 `codespell` 来查找代码中的拼写错误.它也是自动化的.
We also use `codespell` to find typos in code. It is automated as well.
## Metrica B2B 测试 {#metrica-b2b-tests}
## Test Coverage {#test-coverage}
每个 ClickHouse 版本都使用 Yandex Metrica 和 AppMetrica 引擎进行测试. ClickHouse 的测试版和稳定版部署在 VM 上, 并使用 Metrica 引擎的小副本运行, 该引擎处理输入数据的固定样本. 然后将两个 Metrica 引擎实例的结果放在一起比较.
这些测试由单独的团队自动化. 由于移动部件数量众多, 测试在大多数情况下都因完全不相关的原因而失败, 这些原因很难弄清楚. 这些测试很可能对我们有负面价值. 尽管如此, 这些测试在数百次中被证明是有用的.
## 测试覆盖率 {#test-coverage}
我们还跟踪测试覆盖率, 但仅针对功能测试和 clickhouse-server. 它每天进行.
We also track test coverage but only for functional tests and only for clickhouse-server. It is performed on daily basis.
## Tests for Tests
有自动检测薄片测试. 它运行所有新测试100次(用于功能测试)或10次(用于集成测试). 如果至少有一次测试失败，它就被认为是脆弱的.
There is automated check for flaky tests. It runs all new tests 100 times (for functional tests) or 10 times (for integration tests). If at least single time the test failed, it is considered flaky.
## Testflows
[Testflows](https://testflows.com/) 是一个企业级的测试框架. Altinity 使用它进行一些测试, 我们在 CI 中运行这些测试.
[Testflows](https://testflows.com/) is an enterprise-grade open-source testing framework, which is used to test a subset of ClickHouse.
## Yandex 检查 (only for Yandex employees)
## Test Automation {#test-automation}
这些检查将ClickHouse代码导入到Yandex内部的单一存储库中, 所以ClickHouse代码库可以被Yandex的其他产品(YT和YDB)用作库. 请注意, clickhouse-server本身并不是由内部回购构建的, Yandex应用程序使用的是未经修改的开源构建的.
We run tests with [GitHub Actions](https://github.com/features/actions).
## 测试自动化 {#test-automation}
Build jobs and tests are run in Sandbox on per commit basis. Resulting packages and test results are published in GitHub and can be downloaded by direct links. Artifacts are stored for several months. When you send a pull request on GitHub, we tag it as “can be tested” and our CI system will build ClickHouse packages (release, debug, with address sanitizer, etc) for you.
我们使用 Yandex 内部 CI 和名为 "Sandbox" 的作业自动化系统运行测试.
We do not use Travis CI due to the limit on time and computational power.
We do not use Jenkins. It was used before and now we are happy we are not using Jenkins.
在每次提交的基础上, 构建作业和测试都在沙箱中运行. 生成的包和测试结果发布在GitHub上, 可以通过直接链接下载. 产物要保存几个月. 当你在GitHub上发送一个pull请求时, 我们会把它标记为 "可以测试" , 我们的CI系统会为你构建ClickHouse包(发布、调试、使用地址清理器等).
由于时间和计算能力的限制, 我们不使用 Travis CI.
我们不用Jenkins. 以前用过, 现在我们很高兴不用Jenkins了.
[原始文章](https://clickhouse.com/docs/en/development/tests/) <!--hide-->
[Original article](https://clickhouse.com/docs/en/development/tests/) <!--hide-->

3
docs/zh/sql-reference/data-types/lowcardinality.md
View File

@ -55,6 +55,5 @@ ORDER BY id
## 参考
- [高效低基数类型](https://www.altinity.com/blog/2019/3/27/low-cardinality).
- [使用低基数类型减少ClickHouse的存储成本 – 来自Instana工程师的分享](https://www.instana.com/blog/reducing-clickhouse-storage-cost-with-the-low-cardinality-type-lessons-from-an-instana-engineer/).
- [字符优化 (俄语视频分享)](https://youtu.be/rqf-ILRgBdY?list=PL0Z2YDlm0b3iwXCpEFiOOYmwXzVmjJfEt). [英语分享](https://github.com/ClickHouse/clickhouse-presentations/raw/master/meetup19/string_optimization.pdf).
- [字符优化 (俄语视频分享)](https://youtu.be/rqf-ILRgBdY?list=PL0Z2YDlm0b3iwXCpEFiOOYmwXzVmjJfEt). [英语分享](https://github.com/ClickHouse/clickhouse-presentations/raw/master/meetup19/string_optimization.pdf).

2
docs/zh/sql-reference/statements/create.md
View File

@ -121,8 +121,6 @@ ENGINE = <Engine>
...
```
如果指定了编解ec，则默认编解码器不适用。 编解码器可以组合在一个流水线中，例如, `CODEC(Delta, ZSTD)`. 要为您的项目选择最佳的编解码器组合，请通过类似于Altinity中描述的基准测试 [新编码提高ClickHouse效率](https://www.altinity.com/blog/2019/7/new-encodings-to-improve-clickhouse) 文章.
!!! warning "警告"
您无法使用外部实用程序解压缩ClickHouse数据库文件，如 `lz4`. 相反，使用特殊的 [ﾂ环板compressorｮﾂ嘉ｯﾂ偲](https://github.com/ClickHouse/ClickHouse/tree/master/programs/compressor) 实用程序。

67
programs/disks/CommandMkDir.cpp Normal file
View File

@ -0,0 +1,67 @@
#pragma once
#include "ICommand.h"
#include <Interpreters/Context.h>
namespace DB
{
namespace ErrorCodes
{
extern const int BAD_ARGUMENTS;
}
class CommandMkDir : public ICommand
{
public:
CommandMkDir()
{
command_name = "mkdir";
command_option_description.emplace(createOptionsDescription("Allowed options", getTerminalWidth()));
description = "Create directory or directories recursively";
usage = "mkdir [OPTION]... <PATH>";
command_option_description->add_options()
("recursive", "recursively create directories")
;
}
void processOptions(
Poco::Util::LayeredConfiguration & config,
po::variables_map & options) const override
{
if (options.count("recursive"))
config.setBool("recursive", true);
}
void execute(
const std::vector<String> & command_arguments,
DB::ContextMutablePtr & global_context,
Poco::Util::LayeredConfiguration & config) override
{
if (command_arguments.size() != 1)
{
printHelpMessage();
throw DB::Exception("Bad Arguments", DB::ErrorCodes::BAD_ARGUMENTS);
}
String disk_name = config.getString("disk", "default");
String path = command_arguments[0];
DiskPtr disk = global_context->getDisk(disk_name);
String full_path = fullPathWithValidate(disk, path);
bool recursive = config.getBool("recursive", false);
if (recursive)
disk->createDirectories(full_path);
else
disk->createDirectory(full_path);
}
};
}
std::unique_ptr <DB::ICommand> makeCommandMkDir()
{
return std::make_unique<DB::CommandMkDir>();
}

3
programs/disks/DisksApp.cpp
View File

@ -63,7 +63,7 @@ void DisksApp::addOptions(
positional_options_description.add("command_name", 1);
supported_commands = {"list-disks", "list", "move", "remove", "link", "copy", "write", "read"};
supported_commands = {"list-disks", "list", "move", "remove", "link", "copy", "write", "read", "mkdir"};
command_descriptions.emplace("list-disks", makeCommandListDisks());
command_descriptions.emplace("list", makeCommandList());
@ -73,6 +73,7 @@ void DisksApp::addOptions(
command_descriptions.emplace("copy", makeCommandCopy());
command_descriptions.emplace("write", makeCommandWrite());
command_descriptions.emplace("read", makeCommandRead());
command_descriptions.emplace("mkdir", makeCommandMkDir());
}
void DisksApp::processOptions()

1
programs/disks/DisksApp.h
View File

@ -4,6 +4,7 @@
#include "CommandLink.cpp"
#include "CommandList.cpp"
#include "CommandListDisks.cpp"
#include "CommandMkDir.cpp"
#include "CommandMove.cpp"
#include "CommandRead.cpp"
#include "CommandRemove.cpp"

1
programs/disks/ICommand.h
View File

@ -65,3 +65,4 @@ std::unique_ptr <DB::ICommand> makeCommandMove();
std::unique_ptr <DB::ICommand> makeCommandRead();
std::unique_ptr <DB::ICommand> makeCommandRemove();
std::unique_ptr <DB::ICommand> makeCommandWrite();
std::unique_ptr <DB::ICommand> makeCommandMkDir();

56
programs/obfuscator/Obfuscator.cpp
View File

@ -24,6 +24,7 @@
#include <Common/typeid_cast.h>
#include <Common/assert_cast.h>
#include <Formats/registerFormats.h>
#include <Formats/ReadSchemaUtils.h>
#include <Processors/Formats/IInputFormat.h>
#include <QueryPipeline/QueryPipelineBuilder.h>
#include <Processors/Executors/PullingPipelineExecutor.h>
@ -38,6 +39,7 @@
#include <IO/WriteBufferFromFile.h>
#include <Compression/CompressedReadBuffer.h>
#include <Compression/CompressedWriteBuffer.h>
#include <Interpreters/parseColumnsListForTableFunction.h>
#include <memory>
#include <cmath>
#include <unistd.h>
@ -1239,7 +1241,6 @@ try
if (options.count("help")
|| !options.count("seed")
|| !options.count("structure")
|| !options.count("input-format")
|| !options.count("output-format"))
{
@ -1259,7 +1260,11 @@ try
UInt64 seed = sipHash64(options["seed"].as<std::string>());
std::string structure = options["structure"].as<std::string>();
std::string structure;
if (options.count("structure"))
structure = options["structure"].as<std::string>();
std::string input_format = options["input-format"].as<std::string>();
std::string output_format = options["output-format"].as<std::string>();
@ -1287,32 +1292,51 @@ try
markov_model_params.determinator_sliding_window_size = options["determinator-sliding-window-size"].as<UInt64>();
/// Create the header block
std::vector<std::string> structure_vals;
boost::split(structure_vals, structure, boost::algorithm::is_any_of(" ,"), boost::algorithm::token_compress_on);
if (structure_vals.size() % 2 != 0)
throw Exception("Odd number of elements in section structure: must be a list of name type pairs", ErrorCodes::LOGICAL_ERROR);
SharedContextHolder shared_context = Context::createShared();
auto context = Context::createGlobal(shared_context.get());
auto context_const = WithContext(context).getContext();
context->makeGlobalContext();
Block header;
const DataTypeFactory & data_type_factory = DataTypeFactory::instance();
for (size_t i = 0, size = structure_vals.size(); i < size; i += 2)
ColumnsDescription schema_columns;
if (structure.empty())
{
ReadBufferIterator read_buffer_iterator = [&](ColumnsDescription &)
{
auto file = std::make_unique<ReadBufferFromFileDescriptor>(STDIN_FILENO);
/// stdin must be seekable
auto res = lseek(file->getFD(), 0, SEEK_SET);
if (-1 == res)
throwFromErrno("Input must be seekable file (it will be read twice).", ErrorCodes::CANNOT_SEEK_THROUGH_FILE);
return file;
};
schema_columns = readSchemaFromFormat(input_format, {}, read_buffer_iterator, false, context_const);
}
else
{
schema_columns = parseColumnsListFromString(structure, context_const);
}
auto schema_columns_info = schema_columns.getOrdinary();
for (auto & info : schema_columns_info)
{
ColumnWithTypeAndName column;
column.name = structure_vals[i];
column.type = data_type_factory.get(structure_vals[i + 1]);
column.name = info.name;
column.type = info.type;
column.column = column.type->createColumn();
header.insert(std::move(column));
}
SharedContextHolder shared_context = Context::createShared();
auto context = Context::createGlobal(shared_context.get());
context->makeGlobalContext();
ReadBufferFromFileDescriptor file_in(STDIN_FILENO);
WriteBufferFromFileDescriptor file_out(STDOUT_FILENO);
if (load_from_file.empty())
if (load_from_file.empty() || structure.empty())
{
/// stdin must be seekable
auto res = lseek(file_in.getFD(), 0, SEEK_SET);

4
programs/server/Server.cpp
View File

@ -1036,7 +1036,7 @@ int Server::main(const std::vector<std::string> & /*args*/)
try
{
LOG_DEBUG(
log, "Initiailizing merge tree metadata cache lru_cache_size:{} continue_if_corrupted:{}", size, continue_if_corrupted);
log, "Initializing merge tree metadata cache lru_cache_size:{} continue_if_corrupted:{}", size, continue_if_corrupted);
global_context->initializeMergeTreeMetadataCache(path_str + "/" + "rocksdb", size);
}
catch (...)
@ -1089,7 +1089,7 @@ int Server::main(const std::vector<std::string> & /*args*/)
}
}
LOG_DEBUG(log, "Initiailizing interserver credentials.");
LOG_DEBUG(log, "Initializing interserver credentials.");
global_context->updateInterserverCredentials(config());
if (config().has("macros"))

54
src/AggregateFunctions/ThetaSketchData.h
View File

@ -9,6 +9,8 @@
#include <base/StringRef.h>
#include <theta_sketch.hpp>
#include <theta_union.hpp>
#include <theta_intersection.hpp>
#include <theta_a_not_b.hpp>
namespace DB
@ -80,6 +82,58 @@ public:
u->update(rhs.sk_union->get_result());
}
void intersect(const ThetaSketchData & rhs)
{
datasketches::theta_union * u = getSkUnion();
if (sk_update)
{
u->update(*sk_update);
sk_update.reset(nullptr);
}
datasketches::theta_intersection theta_intersection;
theta_intersection.update(u->get_result());
if (rhs.sk_update)
theta_intersection.update(*rhs.sk_update);
else if (rhs.sk_union)
theta_intersection.update(rhs.sk_union->get_result());
sk_union.reset(nullptr);
u = getSkUnion();
u->update(theta_intersection.get_result());
}
void aNotB(const ThetaSketchData & rhs)
{
datasketches::theta_union * u = getSkUnion();
if (sk_update)
{
u->update(*sk_update);
sk_update.reset(nullptr);
}
datasketches::theta_a_not_b a_not_b;
if (rhs.sk_update)
{
datasketches::compact_theta_sketch result = a_not_b.compute(u->get_result(), *rhs.sk_update);
sk_union.reset(nullptr);
u = getSkUnion();
u->update(result);
}
else if (rhs.sk_union)
{
datasketches::compact_theta_sketch result = a_not_b.compute(u->get_result(), rhs.sk_union->get_result());
sk_union.reset(nullptr);
u = getSkUnion();
u->update(result);
}
}
/// You can only call for an empty object.
void read(DB::ReadBuffer & in)
{

2
src/Common/ConcurrencyControl.h
View File

@ -250,7 +250,7 @@ private:
}
}
SlotCount available(std::unique_lock<std::mutex> &)
SlotCount available(std::unique_lock<std::mutex> &) const
{
if (cur_concurrency < max_concurrency)
return max_concurrency - cur_concurrency;

7
src/Common/Config/ConfigProcessor.cpp
View File

@ -13,6 +13,7 @@
#include <Poco/DOM/Text.h>
#include <Poco/DOM/Attr.h>
#include <Poco/DOM/Comment.h>
#include <Poco/XML/XMLWriter.h>
#include <Poco/Util/XMLConfiguration.h>
#include <Common/ZooKeeper/ZooKeeperNodeCache.h>
#include <Common/ZooKeeper/KeeperException.h>
@ -729,7 +730,11 @@ void ConfigProcessor::savePreprocessedConfig(const LoadedConfig & loaded_config,
if (!preprocessed_path_parent.empty())
fs::create_directories(preprocessed_path_parent);
}
DOMWriter().writeNode(preprocessed_path, loaded_config.preprocessed_xml);
DOMWriter writer;
writer.setNewLine("\n");
writer.setIndent(" ");
writer.setOptions(Poco::XML::XMLWriter::PRETTY_PRINT);
writer.writeNode(preprocessed_path, loaded_config.preprocessed_xml);
LOG_DEBUG(log, "Saved preprocessed configuration to '{}'.", preprocessed_path);
}
catch (Poco::Exception & e)

185
src/Common/Config/YAMLParser.cpp
View File

@ -26,114 +26,107 @@ namespace ErrorCodes
extern const int CANNOT_PARSE_YAML;
}
/// A prefix symbol in yaml key
/// We add attributes to nodes by using a prefix symbol in the key part.
/// Currently we use @ as a prefix symbol. Note, that @ is reserved
/// by YAML standard, so we need to write a key-value pair like this: "@attribute": attr_value
const char YAML_ATTRIBUTE_PREFIX = '@';
namespace
{
/// A prefix symbol in yaml key
/// We add attributes to nodes by using a prefix symbol in the key part.
/// Currently we use @ as a prefix symbol. Note, that @ is reserved
/// by YAML standard, so we need to write a key-value pair like this: "@attribute": attr_value
const char YAML_ATTRIBUTE_PREFIX = '@';
Poco::AutoPtr<Poco::XML::Element> createCloneNode(Poco::XML::Element & original_node)
{
Poco::AutoPtr<Poco::XML::Element> clone_node = original_node.ownerDocument()->createElement(original_node.nodeName());
original_node.parentNode()->appendChild(clone_node);
return clone_node;
}
void processNode(const YAML::Node & node, Poco::XML::Element & parent_xml_element)
{
auto * xml_document = parent_xml_element.ownerDocument();
switch (node.Type())
Poco::AutoPtr<Poco::XML::Element> cloneXMLNode(const Poco::XML::Element & original_node)
{
case YAML::NodeType::Scalar:
{
std::string value = node.as<std::string>();
Poco::AutoPtr<Poco::XML::Text> xml_value = xml_document->createTextNode(value);
parent_xml_element.appendChild(xml_value);
break;
}
Poco::AutoPtr<Poco::XML::Element> clone_node = original_node.ownerDocument()->createElement(original_node.nodeName());
original_node.parentNode()->appendChild(clone_node);
return clone_node;
}
/// We process YAML Sequences as a
/// list of <key>value</key> tags with same key and different values.
/// For example, we translate this sequence
/// seq:
/// - val1
/// - val2
///
/// into this:
/// <seq>val1</seq>
/// <seq>val2</seq>
case YAML::NodeType::Sequence:
void processNode(const YAML::Node & node, Poco::XML::Element & parent_xml_node)
{
auto * xml_document = parent_xml_node.ownerDocument();
switch (node.Type())
{
for (const auto & child_node : node)
/// For sequences it depends how we want to process them.
/// Sequences of key-value pairs such as:
/// seq:
/// - k1: val1
/// - k2: val2
/// into xml like this:
/// <seq>
/// <k1>val1</k1>
/// <k2>val2</k2>
/// </seq>
///
/// But, if the sequence is just a list, the root-node needs to be repeated, such as:
/// seq:
/// - val1
/// - val2
/// into xml like this:
/// <seq>val1</seq>
/// <seq>val2</seq>
///
/// Therefore check what type the child is, for further processing.
/// Mixing types (values list or map) will lead to strange results but should not happen.
if (parent_xml_element.hasChildNodes() && !child_node.IsMap())
{
/// Create a new parent node with same tag for each child node
processNode(child_node, *createCloneNode(parent_xml_element));
}
else
{
/// Map, so don't recreate the parent node but add directly
processNode(child_node, parent_xml_element);
}
break;
}
case YAML::NodeType::Map:
{
for (const auto & key_value_pair : node)
case YAML::NodeType::Scalar:
{
const auto & key_node = key_value_pair.first;
const auto & value_node = key_value_pair.second;
std::string key = key_node.as<std::string>();
bool is_attribute = (key.starts_with(YAML_ATTRIBUTE_PREFIX) && value_node.IsScalar());
if (is_attribute)
{
/// we use substr(1) here to remove YAML_ATTRIBUTE_PREFIX from key
auto attribute_name = key.substr(1);
std::string value = value_node.as<std::string>();
parent_xml_element.setAttribute(attribute_name, value);
}
else
{
Poco::AutoPtr<Poco::XML::Element> xml_key = xml_document->createElement(key);
parent_xml_element.appendChild(xml_key);
processNode(value_node, *xml_key);
}
std::string value = node.as<std::string>();
Poco::AutoPtr<Poco::XML::Text> xml_value = xml_document->createTextNode(value);
parent_xml_node.appendChild(xml_value);
break;
}
/// For sequences we repeat the parent xml node. For example,
/// seq:
/// - val1
/// - val2
/// is converted into the following xml:
/// <seq>val1</seq>
/// <seq>val2</seq>
///
/// A sequence of mappings is converted in the same way:
/// seq:
/// - k1: val1
/// k2: val2
/// - k3: val3
/// is converted into the following xml:
/// <seq><k1>val1</k1><k2>val2</k2></seq>
/// <seq><k3>val3</k3></seq>
case YAML::NodeType::Sequence:
{
size_t i = 0;
for (auto it = node.begin(); it != node.end(); ++it, ++i)
{
const auto & child_node = *it;
bool need_clone_parent_xml_node = (i > 0);
if (need_clone_parent_xml_node)
{
/// Create a new parent node with same tag for each child node
processNode(child_node, *cloneXMLNode(parent_xml_node));
}
else
{
/// Map, so don't recreate the parent node but add directly
processNode(child_node, parent_xml_node);
}
}
break;
}
case YAML::NodeType::Map:
{
for (const auto & key_value_pair : node)
{
const auto & key_node = key_value_pair.first;
const auto & value_node = key_value_pair.second;
std::string key = key_node.as<std::string>();
bool is_attribute = (key.starts_with(YAML_ATTRIBUTE_PREFIX) && value_node.IsScalar());
if (is_attribute)
{
/// we use substr(1) here to remove YAML_ATTRIBUTE_PREFIX from key
auto attribute_name = key.substr(1);
std::string value = value_node.as<std::string>();
parent_xml_node.setAttribute(attribute_name, value);
}
else
{
Poco::AutoPtr<Poco::XML::Element> xml_key = xml_document->createElement(key);
parent_xml_node.appendChild(xml_key);
processNode(value_node, *xml_key);
}
}
break;
}
case YAML::NodeType::Null: break;
case YAML::NodeType::Undefined:
{
throw Exception(ErrorCodes::CANNOT_PARSE_YAML, "YAMLParser has encountered node with undefined type and cannot continue parsing of the file");
}
break;
}
case YAML::NodeType::Null: break;
case YAML::NodeType::Undefined:
{
throw Exception(ErrorCodes::CANNOT_PARSE_YAML, "YAMLParser has encountered node with undefined type and cannot continue parsing of the file");
}
}
}
}
Poco::AutoPtr<Poco::XML::Document> YAMLParser::parse(const String& path)
{

7
src/Common/CurrentMemoryTracker.cpp
View File

@ -52,15 +52,10 @@ void CurrentMemoryTracker::allocImpl(Int64 size, bool throw_if_memory_exceeded)
if (current_thread)
{
Int64 will_be = current_thread->untracked_memory + size;
Int64 limit = current_thread->untracked_memory_limit + current_thread->untracked_memory_limit_increase;
if (will_be > limit)
if (will_be > current_thread->untracked_memory_limit)
{
/// Increase limit before track. If tracker throws out-of-limit we would be able to alloc up to untracked_memory_limit bytes
/// more. It could be useful to enlarge Exception message in rethrow logic.
current_thread->untracked_memory_limit_increase = current_thread->untracked_memory_limit;
memory_tracker->allocImpl(will_be, throw_if_memory_exceeded);
current_thread->untracked_memory_limit_increase = 0;
current_thread->untracked_memory = 0;
}
else

12
src/Common/Elf.cpp
View File

@ -22,13 +22,13 @@ Elf::Elf(const std::string & path)
/// Check if it's an elf.
elf_size = in.buffer().size();
if (elf_size < sizeof(ElfEhdr))
throw Exception("The size of supposedly ELF file is too small", ErrorCodes::CANNOT_PARSE_ELF);
throw Exception(ErrorCodes::CANNOT_PARSE_ELF, "The size of supposedly ELF file '{}' is too small", path);
mapped = in.buffer().begin();
header = reinterpret_cast<const ElfEhdr *>(mapped);
if (memcmp(header->e_ident, "\x7F""ELF", 4) != 0)
throw Exception("The file is not ELF according to magic", ErrorCodes::CANNOT_PARSE_ELF);
throw Exception(ErrorCodes::CANNOT_PARSE_ELF, "The file '{}' is not ELF according to magic", path);
/// Get section header.
ElfOff section_header_offset = header->e_shoff;
@ -37,7 +37,7 @@ Elf::Elf(const std::string & path)
if (!section_header_offset
|| !section_header_num_entries
|| section_header_offset + section_header_num_entries * sizeof(ElfShdr) > elf_size)
throw Exception("The ELF is truncated (section header points after end of file)", ErrorCodes::CANNOT_PARSE_ELF);
throw Exception(ErrorCodes::CANNOT_PARSE_ELF, "The ELF '{}' is truncated (section header points after end of file)", path);
section_headers = reinterpret_cast<const ElfShdr *>(mapped + section_header_offset);
@ -48,11 +48,11 @@ Elf::Elf(const std::string & path)
});
if (!section_names_strtab)
throw Exception("The ELF doesn't have string table with section names", ErrorCodes::CANNOT_PARSE_ELF);
throw Exception(ErrorCodes::CANNOT_PARSE_ELF, "The ELF '{}' doesn't have string table with section names", path);
ElfOff section_names_offset = section_names_strtab->header.sh_offset;
if (section_names_offset >= elf_size)
throw Exception("The ELF is truncated (section names string table points after end of file)", ErrorCodes::CANNOT_PARSE_ELF);
throw Exception(ErrorCodes::CANNOT_PARSE_ELF, "The ELF '{}' is truncated (section names string table points after end of file)", path);
section_names = reinterpret_cast<const char *>(mapped + section_names_offset);
@ -64,7 +64,7 @@ Elf::Elf(const std::string & path)
if (!program_header_offset
|| !program_header_num_entries
|| program_header_offset + program_header_num_entries * sizeof(ElfPhdr) > elf_size)
throw Exception("The ELF is truncated (program header points after end of file)", ErrorCodes::CANNOT_PARSE_ELF);
throw Exception(ErrorCodes::CANNOT_PARSE_ELF, "The ELF '{}' is truncated (program header points after end of file)", path);
program_headers = reinterpret_cast<const ElfPhdr *>(mapped + program_header_offset);
}

6
src/Common/FieldVisitorToString.cpp
View File

@ -145,5 +145,11 @@ String FieldVisitorToString::operator() (const Object & x) const
}
String convertFieldToString(const Field & field)
{
if (field.getType() == Field::Types::Which::String)
return field.get<String>();
return applyVisitor(FieldVisitorToString(), field);
}
}

5
src/Common/FieldVisitorToString.h
View File

@ -31,5 +31,8 @@ public:
String operator() (const bool & x) const;
};
}
/// Get value from field and convert it to string.
/// Also remove quotes from strings.
String convertFieldToString(const Field & field);
}

4
src/Common/OvercommitTracker.h
View File

@ -61,7 +61,7 @@ enum class QueryCancellationState
// Usually it's hard to set some reasonable hard memory limit
// (especially, the default value). This class introduces new
// mechanisim for the limiting of memory usage.
// mechanism for the limiting of memory usage.
// Soft limit represents guaranteed amount of memory query/user
// may use. It's allowed to exceed this limit. But if hard limit
// is reached, query with the biggest overcommit ratio
@ -82,7 +82,7 @@ protected:
virtual void pickQueryToExcludeImpl() = 0;
// This mutex is used to disallow concurrent access
// to picked_tracker and cancelation_state variables.
// to picked_tracker and cancellation_state variables.
std::mutex overcommit_m;
std::condition_variable cv;

12
src/Common/ProfileEvents.cpp
View File

@ -286,6 +286,18 @@ The server successfully detected this situation and will download merged part fr
M(S3WriteRequestsThrottling, "Number of 429 and 503 errors in POST, DELETE, PUT and PATCH requests to S3 storage.") \
M(S3WriteRequestsRedirects, "Number of redirects in POST, DELETE, PUT and PATCH requests to S3 storage.") \
\
M(DiskS3ReadMicroseconds, "Time of GET and HEAD requests to DiskS3 storage.") \
M(DiskS3ReadRequestsCount, "Number of GET and HEAD requests to DiskS3 storage.") \
M(DiskS3ReadRequestsErrors, "Number of non-throttling errors in GET and HEAD requests to DiskS3 storage.") \
M(DiskS3ReadRequestsThrottling, "Number of 429 and 503 errors in GET and HEAD requests to DiskS3 storage.") \
M(DiskS3ReadRequestsRedirects, "Number of redirects in GET and HEAD requests to DiskS3 storage.") \
\
M(DiskS3WriteMicroseconds, "Time of POST, DELETE, PUT and PATCH requests to DiskS3 storage.") \
M(DiskS3WriteRequestsCount, "Number of POST, DELETE, PUT and PATCH requests to DiskS3 storage.") \
M(DiskS3WriteRequestsErrors, "Number of non-throttling errors in POST, DELETE, PUT and PATCH requests to DiskS3 storage.") \
M(DiskS3WriteRequestsThrottling, "Number of 429 and 503 errors in POST, DELETE, PUT and PATCH requests to DiskS3 storage.") \
M(DiskS3WriteRequestsRedirects, "Number of redirects in POST, DELETE, PUT and PATCH requests to DiskS3 storage.") \
\
M(ReadBufferFromS3Microseconds, "Time spend in reading from S3.") \
M(ReadBufferFromS3Bytes, "Bytes read from S3.") \
M(ReadBufferFromS3RequestsErrors, "Number of exceptions while reading from S3.") \

1
src/Common/Stopwatch.h
View File

@ -152,4 +152,3 @@ private:
/// Most significant bit is a lock. When it is set, compareAndRestartDeferred method will return false.
UInt64 nanoseconds(UInt64 prev_time) const { return clock_gettime_ns_adjusted(prev_time, clock_type) & 0x7FFFFFFFFFFFFFFFULL; }
};

20
src/Common/SymbolIndex.cpp
View File

@ -37,7 +37,7 @@ But because ClickHouse is linked with most of the symbols exported (-rdynamic fl
It allows to get source file names and line numbers from addresses. Only available if you use -g option for compiler.
It is also used by default for ClickHouse builds, but because of its weight (about two gigabytes)
it is split to separate binary and provided in clickhouse-common-static-dbg package.
This separate binary is placed in /usr/lib/debug/usr/bin/clickhouse and is loaded automatically by tools like gdb, addr2line.
This separate binary is placed in /usr/lib/debug/usr/bin/clickhouse.debug and is loaded automatically by tools like gdb, addr2line.
When you build ClickHouse by yourself, debug info is not split and present in a single huge binary.
What ClickHouse is using to provide good stack traces?
@ -391,10 +391,22 @@ void collectSymbolsFromELF(
std::filesystem::path local_debug_info_path = canonical_path.parent_path() / canonical_path.stem();
local_debug_info_path += ".debug";
std::filesystem::path debug_info_path = std::filesystem::path("/usr/lib/debug") / canonical_path.relative_path();
debug_info_path += ".debug";
if (std::filesystem::exists(local_debug_info_path))
/// NOTE: This is a workaround for current package system.
///
/// Since nfpm cannot copy file only if it exists,
/// and so in cmake empty .debug file is created instead,
/// but if we will try to load empty Elf file, then the CANNOT_PARSE_ELF
/// exception will be thrown from the Elf::Elf.
auto exists_not_empty = [](const std::filesystem::path & path)
{
return std::filesystem::exists(path) && !std::filesystem::is_empty(path);
};
if (exists_not_empty(local_debug_info_path))
object_name = local_debug_info_path;
else if (std::filesystem::exists(debug_info_path))
else if (exists_not_empty(debug_info_path))
object_name = debug_info_path;
else if (build_id.size() >= 2)
{
@ -412,7 +424,7 @@ void collectSymbolsFromELF(
std::filesystem::path build_id_debug_info_path(
fmt::format("/usr/lib/debug/.build-id/{}/{}.debug", build_id_hex.substr(0, 2), build_id_hex.substr(2)));
if (std::filesystem::exists(build_id_debug_info_path))
if (exists_not_empty(build_id_debug_info_path))
object_name = build_id_debug_info_path;
else
object_name = canonical_path;

2
src/Common/ThreadStatus.h
View File

@ -133,8 +133,6 @@ public:
Int64 untracked_memory = 0;
/// Each thread could new/delete memory in range of (-untracked_memory_limit, untracked_memory_limit) without access to common counters.
Int64 untracked_memory_limit = 4 * 1024 * 1024;
/// Increase limit in case of exception.
Int64 untracked_memory_limit_increase = 0;
/// Statistics of read and write rows/bytes
Progress progress_in;

10
src/Common/ZooKeeper/IKeeper.h
View File

@ -80,7 +80,7 @@ enum class Error : int32_t
ZUNIMPLEMENTED = -6, /// Operation is unimplemented
ZOPERATIONTIMEOUT = -7, /// Operation timeout
ZBADARGUMENTS = -8, /// Invalid arguments
ZINVALIDSTATE = -9, /// Invliad zhandle state
ZINVALIDSTATE = -9, /// Invalid zhandle state
/** API errors.
* This is never thrown by the server, it shouldn't be used other than
@ -428,6 +428,12 @@ public:
Exception(const Error code_, const std::string & path); /// NOLINT
Exception(const Exception & exc);
template <typename... Args>
Exception(const Error code_, fmt::format_string<Args...> fmt, Args &&... args)
: Exception(fmt::format(fmt, std::forward<Args>(args)...), code_)
{
}
const char * name() const noexcept override { return "Coordination::Exception"; }
const char * className() const noexcept override { return "Coordination::Exception"; }
Exception * clone() const override { return new Exception(*this); }
@ -439,7 +445,7 @@ public:
/** Usage scenario:
* - create an object and issue commands;
* - you provide callbacks for your commands; callbacks are invoked in internal thread and must be cheap:
* for example, just signal a condvar / fulfull a promise.
* for example, just signal a condvar / fulfill a promise.
* - you also may provide callbacks for watches; they are also invoked in internal thread and must be cheap.
* - whenever you receive exception with ZSESSIONEXPIRED code or method isExpired returns true,
* the ZooKeeper instance is no longer usable - you may only destroy it and probably create another.

18
src/Common/ZooKeeper/TestKeeper.cpp
View File

@ -507,15 +507,15 @@ ResponsePtr TestKeeperSyncRequest::createResponse() const { return std::make_sha
ResponsePtr TestKeeperMultiRequest::createResponse() const { return std::make_shared<MultiResponse>(); }
TestKeeper::TestKeeper(const String & root_path_, Poco::Timespan operation_timeout_)
: root_path(root_path_), operation_timeout(operation_timeout_)
TestKeeper::TestKeeper(const zkutil::ZooKeeperArgs & args_)
: args(args_)
{
container.emplace("/", Node());
if (!root_path.empty())
if (!args.chroot.empty())
{
if (root_path.back() == '/')
root_path.pop_back();
if (args.chroot.back() == '/')
args.chroot.pop_back();
}
processing_thread = ThreadFromGlobalPool([this] { processingThread(); });
@ -547,7 +547,7 @@ void TestKeeper::processingThread()
{
RequestInfo info;
UInt64 max_wait = static_cast<UInt64>(operation_timeout.totalMilliseconds());
UInt64 max_wait = static_cast<UInt64>(args.operation_timeout_ms);
if (requests_queue.tryPop(info, max_wait))
{
if (expired)
@ -556,7 +556,7 @@ void TestKeeper::processingThread()
++zxid;
info.request->addRootPath(root_path);
info.request->addRootPath(args.chroot);
auto [response, _] = info.request->process(container, zxid);
if (info.watch)
@ -580,7 +580,7 @@ void TestKeeper::processingThread()
if (response->error == Error::ZOK)
info.request->processWatches(watches, list_watches);
response->removeRootPath(root_path);
response->removeRootPath(args.chroot);
if (info.callback)
info.callback(*response);
}
@ -689,7 +689,7 @@ void TestKeeper::pushRequest(RequestInfo && request)
if (expired)
throw Exception("Session expired", Error::ZSESSIONEXPIRED);
if (!requests_queue.tryPush(std::move(request), operation_timeout.totalMilliseconds()))
if (!requests_queue.tryPush(std::move(request), args.operation_timeout_ms))
throw Exception("Cannot push request to queue within operation timeout", Error::ZOPERATIONTIMEOUT);
}
catch (...)

8
src/Common/ZooKeeper/TestKeeper.h
View File

@ -8,6 +8,7 @@
#include <Poco/Timespan.h>
#include <Common/ZooKeeper/IKeeper.h>
#include <Common/ZooKeeper/ZooKeeperArgs.h>
#include <Common/ThreadPool.h>
#include <Common/ConcurrentBoundedQueue.h>
@ -33,7 +34,7 @@ using TestKeeperRequestPtr = std::shared_ptr<TestKeeperRequest>;
class TestKeeper final : public IKeeper
{
public:
TestKeeper(const String & root_path_, Poco::Timespan operation_timeout_);
TestKeeper(const zkutil::ZooKeeperArgs & args_);
~TestKeeper() override;
bool isExpired() const override { return expired; }
@ -123,10 +124,7 @@ private:
Container container;
String root_path;
ACLs default_acls;
Poco::Timespan operation_timeout;
zkutil::ZooKeeperArgs args;
std::mutex push_request_mutex;
std::atomic<bool> expired{false};

209
src/Common/ZooKeeper/ZooKeeper.cpp
View File

@ -6,20 +6,18 @@
#include <functional>
#include <filesystem>
#include <Common/randomSeed.h>
#include <base/find_symbols.h>
#include <base/sort.h>
#include <base/getFQDNOrHostName.h>
#include "Common/ZooKeeper/IKeeper.h"
#include <Common/StringUtils/StringUtils.h>
#include <Common/Exception.h>
#include <Common/isLocalAddress.h>
#include <Poco/Net/NetException.h>
#include <Poco/Net/DNS.h>
#define ZOOKEEPER_CONNECTION_TIMEOUT_MS 1000
namespace fs = std::filesystem;
namespace DB
@ -49,25 +47,19 @@ static void check(Coordination::Error code, const std::string & path)
}
void ZooKeeper::init(const std::string & implementation_, const Strings & hosts_, const std::string & identity_,
int32_t session_timeout_ms_, int32_t operation_timeout_ms_, const std::string & chroot_, const GetPriorityForLoadBalancing & get_priority_load_balancing_)
{
log = &Poco::Logger::get("ZooKeeper");
hosts = hosts_;
identity = identity_;
session_timeout_ms = session_timeout_ms_;
operation_timeout_ms = operation_timeout_ms_;
chroot = chroot_;
implementation = implementation_;
get_priority_load_balancing = get_priority_load_balancing_;
void ZooKeeper::init(ZooKeeperArgs args_)
if (implementation == "zookeeper")
{
args = std::move(args_);
log = &Poco::Logger::get("ZooKeeper");
if (args.implementation == "zookeeper")
{
if (hosts.empty())
if (args.hosts.empty())
throw KeeperException("No hosts passed to ZooKeeper constructor.", Coordination::Error::ZBADARGUMENTS);
Coordination::ZooKeeper::Nodes nodes;
nodes.reserve(hosts.size());
nodes.reserve(args.hosts.size());
/// Shuffle the hosts to distribute the load among ZooKeeper nodes.
std::vector<ShuffleHost> shuffled_hosts = shuffleHosts();
@ -108,33 +100,23 @@ void ZooKeeper::init(const std::string & implementation_, const Strings & hosts_
throw KeeperException("Cannot use any of provided ZooKeeper nodes", Coordination::Error::ZBADARGUMENTS);
}
impl = std::make_unique<Coordination::ZooKeeper>(
nodes,
chroot,
identity_.empty() ? "" : "digest",
identity_,
Poco::Timespan(0, session_timeout_ms_ * 1000),
Poco::Timespan(0, ZOOKEEPER_CONNECTION_TIMEOUT_MS * 1000),
Poco::Timespan(0, operation_timeout_ms_ * 1000),
zk_log);
impl = std::make_unique<Coordination::ZooKeeper>(nodes, args, zk_log);
if (chroot.empty())
LOG_TRACE(log, "Initialized, hosts: {}", fmt::join(hosts, ","));
if (args.chroot.empty())
LOG_TRACE(log, "Initialized, hosts: {}", fmt::join(args.hosts, ","));
else
LOG_TRACE(log, "Initialized, hosts: {}, chroot: {}", fmt::join(hosts, ","), chroot);
LOG_TRACE(log, "Initialized, hosts: {}, chroot: {}", fmt::join(args.hosts, ","), args.chroot);
}
else if (implementation == "testkeeper")
else if (args.implementation == "testkeeper")
{
impl = std::make_unique<Coordination::TestKeeper>(
chroot,
Poco::Timespan(0, operation_timeout_ms_ * 1000));
impl = std::make_unique<Coordination::TestKeeper>(args);
}
else
{
throw DB::Exception("Unknown implementation of coordination service: " + implementation, DB::ErrorCodes::NOT_IMPLEMENTED);
throw DB::Exception("Unknown implementation of coordination service: " + args.implementation, DB::ErrorCodes::NOT_IMPLEMENTED);
}
if (!chroot.empty())
if (!args.chroot.empty())
{
/// Here we check that zk root exists.
/// This check is clumsy. The reason is we do this request under common mutex, and never want to hung here.
@ -144,7 +126,7 @@ void ZooKeeper::init(const std::string & implementation_, const Strings & hosts_
/// This should not happen now, when memory tracker is disabled.
/// But let's keep it just in case (it is also easy to backport).
auto future = asyncExists("/");
auto res = future.wait_for(std::chrono::milliseconds(operation_timeout_ms));
auto res = future.wait_for(std::chrono::milliseconds(args.operation_timeout_ms));
if (res != std::future_status::ready)
throw KeeperException("Cannot check if zookeeper root exists.", Coordination::Error::ZOPERATIONTIMEOUT);
@ -153,18 +135,30 @@ void ZooKeeper::init(const std::string & implementation_, const Strings & hosts_
throw KeeperException(code, "/");
if (code == Coordination::Error::ZNONODE)
throw KeeperException("ZooKeeper root doesn't exist. You should create root node " + chroot + " before start.", Coordination::Error::ZNONODE);
throw KeeperException("ZooKeeper root doesn't exist. You should create root node " + args.chroot + " before start.", Coordination::Error::ZNONODE);
}
}
ZooKeeper::ZooKeeper(const ZooKeeperArgs & args_, std::shared_ptr<DB::ZooKeeperLog> zk_log_)
{
zk_log = std::move(zk_log_);
init(args_);
}
ZooKeeper::ZooKeeper(const Poco::Util::AbstractConfiguration & config, const std::string & config_name, std::shared_ptr<DB::ZooKeeperLog> zk_log_)
: zk_log(std::move(zk_log_))
{
init(ZooKeeperArgs(config, config_name));
}
std::vector<ShuffleHost> ZooKeeper::shuffleHosts() const
{
std::function<size_t(size_t index)> get_priority = get_priority_load_balancing.getPriorityFunc(get_priority_load_balancing.load_balancing, 0, hosts.size());
std::function<size_t(size_t index)> get_priority = args.get_priority_load_balancing.getPriorityFunc(args.get_priority_load_balancing.load_balancing, 0, args.hosts.size());
std::vector<ShuffleHost> shuffle_hosts;
for (size_t i = 0; i < hosts.size(); ++i)
for (size_t i = 0; i < args.hosts.size(); ++i)
{
ShuffleHost shuffle_host;
shuffle_host.host = hosts[i];
shuffle_host.host = args.hosts[i];
if (get_priority)
shuffle_host.priority = get_priority(i);
shuffle_host.randomize();
@ -181,125 +175,16 @@ std::vector<ShuffleHost> ZooKeeper::shuffleHosts() const
return shuffle_hosts;
}
ZooKeeper::ZooKeeper(const std::string & hosts_string, const std::string & identity_, int32_t session_timeout_ms_,
int32_t operation_timeout_ms_, const std::string & chroot_, const std::string & implementation_,
std::shared_ptr<DB::ZooKeeperLog> zk_log_, const GetPriorityForLoadBalancing & get_priority_load_balancing_)
{
zk_log = std::move(zk_log_);
Strings hosts_strings;
splitInto<','>(hosts_strings, hosts_string);
init(implementation_, hosts_strings, identity_, session_timeout_ms_, operation_timeout_ms_, chroot_, get_priority_load_balancing_);
}
ZooKeeper::ZooKeeper(const Strings & hosts_, const std::string & identity_, int32_t session_timeout_ms_,
int32_t operation_timeout_ms_, const std::string & chroot_, const std::string & implementation_,
std::shared_ptr<DB::ZooKeeperLog> zk_log_, const GetPriorityForLoadBalancing & get_priority_load_balancing_)
{
zk_log = std::move(zk_log_);
init(implementation_, hosts_, identity_, session_timeout_ms_, operation_timeout_ms_, chroot_, get_priority_load_balancing_);
}
struct ZooKeeperArgs
{
ZooKeeperArgs(const Poco::Util::AbstractConfiguration & config, const std::string & config_name)
{
Poco::Util::AbstractConfiguration::Keys keys;
config.keys(config_name, keys);
session_timeout_ms = Coordination::DEFAULT_SESSION_TIMEOUT_MS;
operation_timeout_ms = Coordination::DEFAULT_OPERATION_TIMEOUT_MS;
implementation = "zookeeper";
for (const auto & key : keys)
{
if (startsWith(key, "node"))
{
hosts.push_back(
(config.getBool(config_name + "." + key + ".secure", false) ? "secure://" : "") +
config.getString(config_name + "." + key + ".host") + ":"
+ config.getString(config_name + "." + key + ".port", "2181")
);
}
else if (key == "session_timeout_ms")
{
session_timeout_ms = config.getInt(config_name + "." + key);
}
else if (key == "operation_timeout_ms")
{
operation_timeout_ms = config.getInt(config_name + "." + key);
}
else if (key == "identity")
{
identity = config.getString(config_name + "." + key);
}
else if (key == "root")
{
chroot = config.getString(config_name + "." + key);
}
else if (key == "implementation")
{
implementation = config.getString(config_name + "." + key);
}
else if (key == "zookeeper_load_balancing")
{
String load_balancing_str = config.getString(config_name + "." + key);
/// Use magic_enum to avoid dependency from dbms (`SettingFieldLoadBalancingTraits::fromString(...)`)
auto load_balancing = magic_enum::enum_cast<DB::LoadBalancing>(Poco::toUpper(load_balancing_str));
if (!load_balancing)
throw DB::Exception(DB::ErrorCodes::BAD_ARGUMENTS, "Unknown load balancing: {}", load_balancing_str);
get_priority_load_balancing.load_balancing = *load_balancing;
}
else
throw KeeperException(std::string("Unknown key ") + key + " in config file", Coordination::Error::ZBADARGUMENTS);
}
if (!chroot.empty())
{
if (chroot.front() != '/')
throw KeeperException(std::string("Root path in config file should start with '/', but got ") + chroot, Coordination::Error::ZBADARGUMENTS);
if (chroot.back() == '/')
chroot.pop_back();
}
/// init get_priority_load_balancing
get_priority_load_balancing.hostname_differences.resize(hosts.size());
const String & local_hostname = getFQDNOrHostName();
for (size_t i = 0; i < hosts.size(); ++i)
{
const String & node_host = hosts[i].substr(0, hosts[i].find_last_of(':'));
get_priority_load_balancing.hostname_differences[i] = DB::getHostNameDifference(local_hostname, node_host);
}
}
Strings hosts;
std::string identity;
int session_timeout_ms;
int operation_timeout_ms;
std::string chroot;
std::string implementation;
GetPriorityForLoadBalancing get_priority_load_balancing;
};
ZooKeeper::ZooKeeper(const Poco::Util::AbstractConfiguration & config, const std::string & config_name, std::shared_ptr<DB::ZooKeeperLog> zk_log_)
: zk_log(std::move(zk_log_))
{
ZooKeeperArgs args(config, config_name);
init(args.implementation, args.hosts, args.identity, args.session_timeout_ms, args.operation_timeout_ms, args.chroot, args.get_priority_load_balancing);
}
bool ZooKeeper::configChanged(const Poco::Util::AbstractConfiguration & config, const std::string & config_name) const
{
ZooKeeperArgs args(config, config_name);
ZooKeeperArgs new_args(config, config_name);
// skip reload testkeeper cause it's for test and data in memory
if (args.implementation == implementation && implementation == "testkeeper")
if (new_args.implementation == args.implementation && args.implementation == "testkeeper")
return false;
if (args.get_priority_load_balancing != get_priority_load_balancing)
return true;
return std::tie(args.implementation, args.hosts, args.identity, args.session_timeout_ms, args.operation_timeout_ms, args.chroot, args.get_priority_load_balancing)
!= std::tie(implementation, hosts, identity, session_timeout_ms, operation_timeout_ms, chroot, args.get_priority_load_balancing);
return args != new_args;
}
@ -318,7 +203,7 @@ Coordination::Error ZooKeeper::getChildrenImpl(const std::string & path, Strings
{
auto future_result = asyncTryGetChildrenNoThrow(path, watch_callback, list_request_type);
if (future_result.wait_for(std::chrono::milliseconds(operation_timeout_ms)) != std::future_status::ready)
if (future_result.wait_for(std::chrono::milliseconds(args.operation_timeout_ms)) != std::future_status::ready)
{
impl->finalize(fmt::format("Operation timeout on {} {}", toString(Coordination::OpNum::List), path));
return Coordination::Error::ZOPERATIONTIMEOUT;
@ -385,7 +270,7 @@ Coordination::Error ZooKeeper::createImpl(const std::string & path, const std::s
{
auto future_result = asyncTryCreateNoThrow(path, data, mode);
if (future_result.wait_for(std::chrono::milliseconds(operation_timeout_ms)) != std::future_status::ready)
if (future_result.wait_for(std::chrono::milliseconds(args.operation_timeout_ms)) != std::future_status::ready)
{
impl->finalize(fmt::format("Operation timeout on {} {}", toString(Coordination::OpNum::Create), path));
return Coordination::Error::ZOPERATIONTIMEOUT;
@ -455,7 +340,7 @@ Coordination::Error ZooKeeper::removeImpl(const std::string & path, int32_t vers
auto future_result = asyncTryRemoveNoThrow(path, version);
if (future_result.wait_for(std::chrono::milliseconds(operation_timeout_ms)) != std::future_status::ready)
if (future_result.wait_for(std::chrono::milliseconds(args.operation_timeout_ms)) != std::future_status::ready)
{
impl->finalize(fmt::format("Operation timeout on {} {}", toString(Coordination::OpNum::Remove), path));
return Coordination::Error::ZOPERATIONTIMEOUT;
@ -487,7 +372,7 @@ Coordination::Error ZooKeeper::existsImpl(const std::string & path, Coordination
{
auto future_result = asyncTryExistsNoThrow(path, watch_callback);
if (future_result.wait_for(std::chrono::milliseconds(operation_timeout_ms)) != std::future_status::ready)
if (future_result.wait_for(std::chrono::milliseconds(args.operation_timeout_ms)) != std::future_status::ready)
{
impl->finalize(fmt::format("Operation timeout on {} {}", toString(Coordination::OpNum::Exists), path));
return Coordination::Error::ZOPERATIONTIMEOUT;
@ -521,7 +406,7 @@ Coordination::Error ZooKeeper::getImpl(const std::string & path, std::string & r
{
auto future_result = asyncTryGetNoThrow(path, watch_callback);
if (future_result.wait_for(std::chrono::milliseconds(operation_timeout_ms)) != std::future_status::ready)
if (future_result.wait_for(std::chrono::milliseconds(args.operation_timeout_ms)) != std::future_status::ready)
{
impl->finalize(fmt::format("Operation timeout on {} {}", toString(Coordination::OpNum::Get), path));
return Coordination::Error::ZOPERATIONTIMEOUT;
@ -593,7 +478,7 @@ Coordination::Error ZooKeeper::setImpl(const std::string & path, const std::stri
{
auto future_result = asyncTrySetNoThrow(path, data, version);
if (future_result.wait_for(std::chrono::milliseconds(operation_timeout_ms)) != std::future_status::ready)
if (future_result.wait_for(std::chrono::milliseconds(args.operation_timeout_ms)) != std::future_status::ready)
{
impl->finalize(fmt::format("Operation timeout on {} {}", toString(Coordination::OpNum::Set), path));
return Coordination::Error::ZOPERATIONTIMEOUT;
@ -645,7 +530,7 @@ Coordination::Error ZooKeeper::multiImpl(const Coordination::Requests & requests
auto future_result = asyncTryMultiNoThrow(requests);
if (future_result.wait_for(std::chrono::milliseconds(operation_timeout_ms)) != std::future_status::ready)
if (future_result.wait_for(std::chrono::milliseconds(args.operation_timeout_ms)) != std::future_status::ready)
{
impl->finalize(fmt::format("Operation timeout on {} {}", toString(Coordination::OpNum::Multi), requests[0]->getPath()));
return Coordination::Error::ZOPERATIONTIMEOUT;
@ -679,7 +564,7 @@ Coordination::Error ZooKeeper::syncImpl(const std::string & path, std::string &
{
auto future_result = asyncTrySyncNoThrow(path);
if (future_result.wait_for(std::chrono::milliseconds(operation_timeout_ms)) != std::future_status::ready)
if (future_result.wait_for(std::chrono::milliseconds(args.operation_timeout_ms)) != std::future_status::ready)
{
impl->finalize(fmt::format("Operation timeout on {} {}", toString(Coordination::OpNum::Sync), path));
return Coordination::Error::ZOPERATIONTIMEOUT;
@ -884,7 +769,7 @@ void ZooKeeper::waitForEphemeralToDisappearIfAny(const std::string & path)
if (!tryGet(path, content, nullptr, eph_node_disappeared))
return;
int32_t timeout_ms = 3 * session_timeout_ms;
int32_t timeout_ms = 3 * args.session_timeout_ms;
if (!eph_node_disappeared->tryWait(timeout_ms))
throw DB::Exception(DB::ErrorCodes::LOGICAL_ERROR,
"Ephemeral node {} still exists after {}s, probably it's owned by someone else. "
@ -894,7 +779,7 @@ void ZooKeeper::waitForEphemeralToDisappearIfAny(const std::string & path)
ZooKeeperPtr ZooKeeper::startNewSession() const
{
return std::make_shared<ZooKeeper>(hosts, identity, session_timeout_ms, operation_timeout_ms, chroot, implementation, zk_log, get_priority_load_balancing);
return std::make_shared<ZooKeeper>(args, zk_log);
}

31
src/Common/ZooKeeper/ZooKeeper.h
View File

@ -13,7 +13,7 @@
#include <Common/Stopwatch.h>
#include <Common/ZooKeeper/IKeeper.h>
#include <Common/ZooKeeper/ZooKeeperConstants.h>
#include <Common/GetPriorityForLoadBalancing.h>
#include <Common/ZooKeeper/ZooKeeperArgs.h>
#include <Common/thread_local_rng.h>
#include <unistd.h>
#include <random>
@ -72,24 +72,11 @@ using GetPriorityForLoadBalancing = DB::GetPriorityForLoadBalancing;
class ZooKeeper
{
public:
using Ptr = std::shared_ptr<ZooKeeper>;
/// hosts_string -- comma separated [secure://]host:port list
explicit ZooKeeper(const std::string & hosts_string, const std::string & identity_ = "",
int32_t session_timeout_ms_ = Coordination::DEFAULT_SESSION_TIMEOUT_MS,
int32_t operation_timeout_ms_ = Coordination::DEFAULT_OPERATION_TIMEOUT_MS,
const std::string & chroot_ = "",
const std::string & implementation_ = "zookeeper",
std::shared_ptr<DB::ZooKeeperLog> zk_log_ = nullptr,
const GetPriorityForLoadBalancing & get_priority_load_balancing_ = {});
ZooKeeper(const ZooKeeperArgs & args_, std::shared_ptr<DB::ZooKeeperLog> zk_log_ = nullptr);
explicit ZooKeeper(const Strings & hosts_, const std::string & identity_ = "",
int32_t session_timeout_ms_ = Coordination::DEFAULT_SESSION_TIMEOUT_MS,
int32_t operation_timeout_ms_ = Coordination::DEFAULT_OPERATION_TIMEOUT_MS,
const std::string & chroot_ = "",
const std::string & implementation_ = "zookeeper",
std::shared_ptr<DB::ZooKeeperLog> zk_log_ = nullptr,
const GetPriorityForLoadBalancing & get_priority_load_balancing_ = {});
/** Config of the form:
<zookeeper>
@ -337,8 +324,7 @@ public:
private:
friend class EphemeralNodeHolder;
void init(const std::string & implementation_, const Strings & hosts_, const std::string & identity_,
int32_t session_timeout_ms_, int32_t operation_timeout_ms_, const std::string & chroot_, const GetPriorityForLoadBalancing & get_priority_load_balancing_);
void init(ZooKeeperArgs args_);
/// The following methods don't any throw exceptions but return error codes.
Coordination::Error createImpl(const std::string & path, const std::string & data, int32_t mode, std::string & path_created);
@ -358,20 +344,13 @@ private:
std::unique_ptr<Coordination::IKeeper> impl;
Strings hosts;
std::string identity;
int32_t session_timeout_ms;
int32_t operation_timeout_ms;
std::string chroot;
std::string implementation;
ZooKeeperArgs args;
std::mutex mutex;
Poco::Logger * log = nullptr;
std::shared_ptr<DB::ZooKeeperLog> zk_log;
GetPriorityForLoadBalancing get_priority_load_balancing;
AtomicStopwatch session_uptime;
};

108
src/Common/ZooKeeper/ZooKeeperArgs.cpp Normal file
View File

@ -0,0 +1,108 @@
#include <Common/ZooKeeper/ZooKeeperArgs.h>
#include <Common/ZooKeeper/KeeperException.h>
#include <base/find_symbols.h>
#include <base/getFQDNOrHostName.h>
#include <Poco/Util/AbstractConfiguration.h>
#include <Common/isLocalAddress.h>
#include <Poco/String.h>
namespace DB
{
namespace ErrorCodes
{
extern const int BAD_ARGUMENTS;
}
}
namespace zkutil
{
ZooKeeperArgs::ZooKeeperArgs(const Poco::Util::AbstractConfiguration & config, const String & config_name)
{
Poco::Util::AbstractConfiguration::Keys keys;
config.keys(config_name, keys);
for (const auto & key : keys)
{
if (key.starts_with("node"))
{
hosts.push_back(
(config.getBool(config_name + "." + key + ".secure", false) ? "secure://" : "")
+ config.getString(config_name + "." + key + ".host") + ":" + config.getString(config_name + "." + key + ".port", "2181"));
}
else if (key == "session_timeout_ms")
{
session_timeout_ms = config.getInt(config_name + "." + key);
}
else if (key == "operation_timeout_ms")
{
operation_timeout_ms = config.getInt(config_name + "." + key);
}
else if (key == "connection_timeout_ms")
{
connection_timeout_ms = config.getInt(config_name + "." + key);
}
else if (key == "send_fault_probability")
{
send_fault_probability = config.getDouble(config_name + "." + key);
}
else if (key == "recv_fault_probability")
{
recv_fault_probability = config.getDouble(config_name + "." + key);
}
else if (key == "identity")
{
identity = config.getString(config_name + "." + key);
if (!identity.empty())
auth_scheme = "digest";
}
else if (key == "root")
{
chroot = config.getString(config_name + "." + key);
}
else if (key == "implementation")
{
implementation = config.getString(config_name + "." + key);
}
else if (key == "zookeeper_load_balancing")
{
String load_balancing_str = config.getString(config_name + "." + key);
/// Use magic_enum to avoid dependency from dbms (`SettingFieldLoadBalancingTraits::fromString(...)`)
auto load_balancing = magic_enum::enum_cast<DB::LoadBalancing>(Poco::toUpper(load_balancing_str));
if (!load_balancing)
throw DB::Exception(DB::ErrorCodes::BAD_ARGUMENTS, "Unknown load balancing: {}", load_balancing_str);
get_priority_load_balancing.load_balancing = *load_balancing;
}
else
throw KeeperException(std::string("Unknown key ") + key + " in config file", Coordination::Error::ZBADARGUMENTS);
}
if (!chroot.empty())
{
if (chroot.front() != '/')
throw KeeperException(
Coordination::Error::ZBADARGUMENTS,
"Root path in config file should start with '/', but got {}", chroot);
if (chroot.back() == '/')
chroot.pop_back();
}
if (session_timeout_ms < 0 || operation_timeout_ms < 0 || connection_timeout_ms < 0)
throw KeeperException("Timeout cannot be negative", Coordination::Error::ZBADARGUMENTS);
/// init get_priority_load_balancing
get_priority_load_balancing.hostname_differences.resize(hosts.size());
const String & local_hostname = getFQDNOrHostName();
for (size_t i = 0; i < hosts.size(); ++i)
{
const String & node_host = hosts[i].substr(0, hosts[i].find_last_of(':'));
get_priority_load_balancing.hostname_differences[i] = DB::getHostNameDifference(local_hostname, node_host);
}
}
ZooKeeperArgs::ZooKeeperArgs(const String & hosts_string)
{
splitInto<','>(hosts, hosts_string);
}
}

37
src/Common/ZooKeeper/ZooKeeperArgs.h Normal file
View File

@ -0,0 +1,37 @@
#pragma once
#include <Common/ZooKeeper/Types.h>
#include <Common/ZooKeeper/ZooKeeperConstants.h>
#include <Common/GetPriorityForLoadBalancing.h>
namespace Poco::Util
{
class AbstractConfiguration;
}
namespace zkutil
{
struct ZooKeeperArgs
{
ZooKeeperArgs(const Poco::Util::AbstractConfiguration & config, const String & config_name);
/// hosts_string -- comma separated [secure://]host:port list
ZooKeeperArgs(const String & hosts_string);
ZooKeeperArgs() = default;
bool operator == (const ZooKeeperArgs &) const = default;
String implementation = "zookeeper";
Strings hosts;
String auth_scheme;
String identity;
String chroot;
int32_t connection_timeout_ms = Coordination::DEFAULT_CONNECTION_TIMEOUT_MS;
int32_t session_timeout_ms = Coordination::DEFAULT_SESSION_TIMEOUT_MS;
int32_t operation_timeout_ms = Coordination::DEFAULT_OPERATION_TIMEOUT_MS;
float send_fault_probability = 0;
float recv_fault_probability = 0;
DB::GetPriorityForLoadBalancing get_priority_load_balancing;
};
}

1
src/Common/ZooKeeper/ZooKeeperConstants.h
View File

@ -56,5 +56,6 @@ static constexpr int32_t DEFAULT_SESSION_TIMEOUT_MS = 30000;
static constexpr int32_t DEFAULT_MIN_SESSION_TIMEOUT_MS = 10000;
static constexpr int32_t DEFAULT_MAX_SESSION_TIMEOUT_MS = 100000;
static constexpr int32_t DEFAULT_OPERATION_TIMEOUT_MS = 10000;
static constexpr int32_t DEFAULT_CONNECTION_TIMEOUT_MS = 1000;
}

141
src/Common/ZooKeeper/ZooKeeperImpl.cpp
View File

@ -276,15 +276,15 @@ void ZooKeeper::read(T & x)
Coordination::read(x, *in);
}
static void removeRootPath(String & path, const String & root_path)
static void removeRootPath(String & path, const String & chroot)
{
if (root_path.empty())
if (chroot.empty())
return;
if (path.size() <= root_path.size())
throw Exception("Received path is not longer than root_path", Error::ZDATAINCONSISTENCY);
if (path.size() <= chroot.size())
throw Exception(Error::ZDATAINCONSISTENCY, "Received path is not longer than chroot");
path = path.substr(root_path.size());
path = path.substr(chroot.size());
}
ZooKeeper::~ZooKeeper()
@ -308,27 +308,20 @@ ZooKeeper::~ZooKeeper()
ZooKeeper::ZooKeeper(
const Nodes & nodes,
const String & root_path_,
const String & auth_scheme,
const String & auth_data,
Poco::Timespan session_timeout_,
Poco::Timespan connection_timeout,
Poco::Timespan operation_timeout_,
const zkutil::ZooKeeperArgs & args_,
std::shared_ptr<ZooKeeperLog> zk_log_)
: root_path(root_path_),
session_timeout(session_timeout_),
operation_timeout(std::min(operation_timeout_, session_timeout_))
: args(args_)
{
log = &Poco::Logger::get("ZooKeeperClient");
std::atomic_store(&zk_log, std::move(zk_log_));
if (!root_path.empty())
if (!args.chroot.empty())
{
if (root_path.back() == '/')
root_path.pop_back();
if (args.chroot.back() == '/')
args.chroot.pop_back();
}
if (auth_scheme.empty())
if (args.auth_scheme.empty())
{
ACL acl;
acl.permissions = ACL::All;
@ -345,10 +338,22 @@ ZooKeeper::ZooKeeper(
default_acls.emplace_back(std::move(acl));
}
connect(nodes, connection_timeout);
if (!auth_scheme.empty())
sendAuth(auth_scheme, auth_data);
/// It makes sense (especially, for async requests) to inject a fault in two places:
/// pushRequest (before request is sent) and receiveEvent (after request was executed).
if (0 < args.send_fault_probability && args.send_fault_probability <= 1)
{
send_inject_fault.emplace(args.send_fault_probability);
}
if (0 < args.recv_fault_probability && args.recv_fault_probability <= 1)
{
recv_inject_fault.emplace(args.recv_fault_probability);
}
connect(nodes, args.connection_timeout_ms * 1000);
if (!args.auth_scheme.empty())
sendAuth(args.auth_scheme, args.identity);
send_thread = ThreadFromGlobalPool([this] { sendThread(); });
receive_thread = ThreadFromGlobalPool([this] { receiveThread(); });
@ -364,7 +369,7 @@ void ZooKeeper::connect(
Poco::Timespan connection_timeout)
{
if (nodes.empty())
throw Exception("No nodes passed to ZooKeeper constructor", Error::ZBADARGUMENTS);
throw Exception(Error::ZBADARGUMENTS, "No nodes passed to ZooKeeper constructor");
static constexpr size_t num_tries = 3;
bool connected = false;
@ -394,8 +399,8 @@ void ZooKeeper::connect(
socket.connect(node.address, connection_timeout);
socket_address = socket.peerAddress();
socket.setReceiveTimeout(operation_timeout);
socket.setSendTimeout(operation_timeout);
socket.setReceiveTimeout(args.operation_timeout_ms * 1000);
socket.setSendTimeout(args.operation_timeout_ms * 1000);
socket.setNoDelay(true);
in.emplace(socket);
@ -453,7 +458,7 @@ void ZooKeeper::connect(
}
message << fail_reasons.str() << "\n";
throw Exception(message.str(), Error::ZCONNECTIONLOSS);
throw Exception(Error::ZCONNECTIONLOSS, message.str());
}
else
{
@ -466,7 +471,7 @@ void ZooKeeper::sendHandshake()
{
int32_t handshake_length = 44;
int64_t last_zxid_seen = 0;
int32_t timeout = session_timeout.totalMilliseconds();
int32_t timeout = args.session_timeout_ms;
int64_t previous_session_id = 0; /// We don't support session restore. So previous session_id is always zero.
constexpr int32_t passwd_len = 16;
std::array<char, passwd_len> passwd {};
@ -491,7 +496,7 @@ void ZooKeeper::receiveHandshake()
read(handshake_length);
if (handshake_length != SERVER_HANDSHAKE_LENGTH)
throw Exception("Unexpected handshake length received: " + DB::toString(handshake_length), Error::ZMARSHALLINGERROR);
throw Exception(Error::ZMARSHALLINGERROR, "Unexpected handshake length received: {}", handshake_length);
read(protocol_version_read);
if (protocol_version_read != ZOOKEEPER_PROTOCOL_VERSION)
@ -500,15 +505,15 @@ void ZooKeeper::receiveHandshake()
/// It's better for faster failover than just connection drop.
/// Implemented in clickhouse-keeper.
if (protocol_version_read == KEEPER_PROTOCOL_VERSION_CONNECTION_REJECT)
throw Exception("Keeper server rejected the connection during the handshake. Possibly it's overloaded, doesn't see leader or stale", Error::ZCONNECTIONLOSS);
throw Exception(Error::ZCONNECTIONLOSS, "Keeper server rejected the connection during the handshake. Possibly it's overloaded, doesn't see leader or stale");
else
throw Exception("Unexpected protocol version: " + DB::toString(protocol_version_read), Error::ZMARSHALLINGERROR);
throw Exception(Error::ZMARSHALLINGERROR, "Unexpected protocol version: {}", protocol_version_read);
}
read(timeout);
if (timeout != session_timeout.totalMilliseconds())
if (timeout != args.session_timeout_ms)
/// Use timeout from server.
session_timeout = timeout * Poco::Timespan::MILLISECONDS;
args.session_timeout_ms = timeout;
read(session_id);
read(passwd);
@ -535,17 +540,15 @@ void ZooKeeper::sendAuth(const String & scheme, const String & data)
read(err);
if (read_xid != AUTH_XID)
throw Exception("Unexpected event received in reply to auth request: " + DB::toString(read_xid),
Error::ZMARSHALLINGERROR);
throw Exception(Error::ZMARSHALLINGERROR, "Unexpected event received in reply to auth request: {}", read_xid);
int32_t actual_length = in->count() - count_before_event;
if (length != actual_length)
throw Exception("Response length doesn't match. Expected: " + DB::toString(length) + ", actual: " + DB::toString(actual_length),
Error::ZMARSHALLINGERROR);
throw Exception(Error::ZMARSHALLINGERROR, "Response length doesn't match. Expected: {}, actual: {}", length, actual_length);
if (err != Error::ZOK)
throw Exception("Error received in reply to auth request. Code: " + DB::toString(static_cast<int32_t>(err)) + ". Message: " + String(errorMessage(err)),
Error::ZMARSHALLINGERROR);
throw Exception(Error::ZMARSHALLINGERROR, "Error received in reply to auth request. Code: {}. Message: {}",
static_cast<int32_t>(err), errorMessage(err));
}
@ -562,14 +565,14 @@ void ZooKeeper::sendThread()
auto prev_bytes_sent = out->count();
auto now = clock::now();
auto next_heartbeat_time = prev_heartbeat_time + std::chrono::milliseconds(session_timeout.totalMilliseconds() / 3);
auto next_heartbeat_time = prev_heartbeat_time + std::chrono::milliseconds(args.session_timeout_ms / 3);
if (next_heartbeat_time > now)
{
/// Wait for the next request in queue. No more than operation timeout. No more than until next heartbeat time.
UInt64 max_wait = std::min(
static_cast<UInt64>(std::chrono::duration_cast<std::chrono::milliseconds>(next_heartbeat_time - now).count()),
static_cast<UInt64>(operation_timeout.totalMilliseconds()));
static_cast<UInt64>(args.operation_timeout_ms));
RequestInfo info;
if (requests_queue.tryPop(info, max_wait))
@ -594,7 +597,7 @@ void ZooKeeper::sendThread()
break;
}
info.request->addRootPath(root_path);
info.request->addRootPath(args.chroot);
info.request->probably_sent = true;
info.request->write(*out);
@ -633,13 +636,13 @@ void ZooKeeper::receiveThread()
try
{
Int64 waited = 0;
Int64 waited_us = 0;
while (!requests_queue.isFinished())
{
auto prev_bytes_received = in->count();
clock::time_point now = clock::now();
UInt64 max_wait = operation_timeout.totalMicroseconds();
UInt64 max_wait_us = args.operation_timeout_ms * 1000;
std::optional<RequestInfo> earliest_operation;
{
@ -648,30 +651,32 @@ void ZooKeeper::receiveThread()
{
/// Operations are ordered by xid (and consequently, by time).
earliest_operation = operations.begin()->second;
auto earliest_operation_deadline = earliest_operation->time + std::chrono::microseconds(operation_timeout.totalMicroseconds());
auto earliest_operation_deadline = earliest_operation->time + std::chrono::microseconds(args.operation_timeout_ms * 1000);
if (now > earliest_operation_deadline)
throw Exception("Operation timeout (deadline already expired) for path: " + earliest_operation->request->getPath(), Error::ZOPERATIONTIMEOUT);
max_wait = std::chrono::duration_cast<std::chrono::microseconds>(earliest_operation_deadline - now).count();
throw Exception(Error::ZOPERATIONTIMEOUT, "Operation timeout (deadline already expired) for path: {}",
earliest_operation->request->getPath());
max_wait_us = std::chrono::duration_cast<std::chrono::microseconds>(earliest_operation_deadline - now).count();
}
}
if (in->poll(max_wait))
if (in->poll(max_wait_us))
{
if (requests_queue.isFinished())
break;
receiveEvent();
waited = 0;
waited_us = 0;
}
else
{
if (earliest_operation)
{
throw Exception("Operation timeout (no response) for request " + toString(earliest_operation->request->getOpNum()) + " for path: " + earliest_operation->request->getPath(), Error::ZOPERATIONTIMEOUT);
throw Exception(Error::ZOPERATIONTIMEOUT, "Operation timeout (no response) for request {} for path: {}",
earliest_operation->request->getOpNum(), earliest_operation->request->getPath());
}
waited += max_wait;
if (waited >= session_timeout.totalMicroseconds())
throw Exception("Nothing is received in session timeout", Error::ZOPERATIONTIMEOUT);
waited_us += max_wait_us;
if (waited_us >= args.session_timeout_ms * 1000)
throw Exception(Error::ZOPERATIONTIMEOUT, "Nothing is received in session timeout");
}
@ -703,10 +708,13 @@ void ZooKeeper::receiveEvent()
ZooKeeperResponsePtr response;
UInt64 elapsed_ms = 0;
if (unlikely(recv_inject_fault) && recv_inject_fault.value()(thread_local_rng))
throw Exception(Error::ZSESSIONEXPIRED, "Session expired (fault injected on recv)");
if (xid == PING_XID)
{
if (err != Error::ZOK)
throw Exception("Received error in heartbeat response: " + String(errorMessage(err)), Error::ZRUNTIMEINCONSISTENCY);
throw Exception(Error::ZRUNTIMEINCONSISTENCY, "Received error in heartbeat response: {}", errorMessage(err));
response = std::make_shared<ZooKeeperHeartbeatResponse>();
}
@ -781,7 +789,7 @@ void ZooKeeper::receiveEvent()
else
{
response->readImpl(*in);
response->removeRootPath(root_path);
response->removeRootPath(args.chroot);
}
/// Instead of setting the watch in sendEvent, set it in receiveEvent because need to check the response.
/// The watch shouldn't be set if the node does not exist and it will never exist like sequential ephemeral nodes.
@ -801,9 +809,9 @@ void ZooKeeper::receiveEvent()
{
CurrentMetrics::add(CurrentMetrics::ZooKeeperWatch);
/// The key of wathces should exclude the root_path
/// The key of wathces should exclude the args.chroot
String req_path = request_info.request->getPath();
removeRootPath(req_path, root_path);
removeRootPath(req_path, args.chroot);
std::lock_guard lock(watches_mutex);
watches[req_path].emplace_back(std::move(request_info.watch));
}
@ -811,7 +819,7 @@ void ZooKeeper::receiveEvent()
int32_t actual_length = in->count() - count_before_event;
if (length != actual_length)
throw Exception("Response length doesn't match. Expected: " + DB::toString(length) + ", actual: " + DB::toString(actual_length), Error::ZMARSHALLINGERROR);
throw Exception(Error::ZMARSHALLINGERROR, "Response length doesn't match. Expected: {}, actual: {}", length, actual_length);
logOperationIfNeeded(request_info.request, response, /* finalize= */ false, elapsed_ms); //-V614
}
@ -1035,9 +1043,9 @@ void ZooKeeper::pushRequest(RequestInfo && info)
{
info.request->xid = next_xid.fetch_add(1);
if (info.request->xid == CLOSE_XID)
throw Exception("xid equal to close_xid", Error::ZSESSIONEXPIRED);
throw Exception(Error::ZSESSIONEXPIRED, "xid equal to close_xid");
if (info.request->xid < 0)
throw Exception("XID overflow", Error::ZSESSIONEXPIRED);
throw Exception(Error::ZSESSIONEXPIRED, "XID overflow");
if (auto * multi_request = dynamic_cast<ZooKeeperMultiRequest *>(info.request.get()))
{
@ -1046,12 +1054,15 @@ void ZooKeeper::pushRequest(RequestInfo && info)
}
}
if (!requests_queue.tryPush(std::move(info), operation_timeout.totalMilliseconds()))
if (unlikely(send_inject_fault) && send_inject_fault.value()(thread_local_rng))
throw Exception(Error::ZSESSIONEXPIRED, "Session expired (fault injected on send)");
if (!requests_queue.tryPush(std::move(info), args.operation_timeout_ms))
{
if (requests_queue.isFinished())
throw Exception("Session expired", Error::ZSESSIONEXPIRED);
throw Exception(Error::ZSESSIONEXPIRED, "Session expired");
throw Exception("Cannot push request to queue within operation timeout", Error::ZOPERATIONTIMEOUT);
throw Exception(Error::ZOPERATIONTIMEOUT, "Cannot push request to queue within operation timeout");
}
}
catch (...)
@ -1079,7 +1090,7 @@ void ZooKeeper::initApiVersion()
};
get(keeper_api_version_path, std::move(callback), {});
if (future.wait_for(std::chrono::milliseconds(operation_timeout.totalMilliseconds())) != std::future_status::ready)
if (future.wait_for(std::chrono::milliseconds(args.operation_timeout_ms)) != std::future_status::ready)
{
LOG_TRACE(log, "Failed to get API version: timeout");
return;
@ -1220,7 +1231,7 @@ void ZooKeeper::list(
if (keeper_api_version < Coordination::KeeperApiVersion::WITH_FILTERED_LIST)
{
if (list_request_type != ListRequestType::ALL)
throw Exception("Filtered list request type cannot be used because it's not supported by the server", Error::ZBADARGUMENTS);
throw Exception(Error::ZBADARGUMENTS, "Filtered list request type cannot be used because it's not supported by the server");
request = std::make_shared<ZooKeeperListRequest>();
}
@ -1299,8 +1310,8 @@ void ZooKeeper::close()
RequestInfo request_info;
request_info.request = std::make_shared<ZooKeeperCloseRequest>(std::move(request));
if (!requests_queue.tryPush(std::move(request_info), operation_timeout.totalMilliseconds()))
throw Exception("Cannot push close request to queue within operation timeout", Error::ZOPERATIONTIMEOUT);
if (!requests_queue.tryPush(std::move(request_info), args.operation_timeout_ms))
throw Exception(Error::ZOPERATIONTIMEOUT, "Cannot push close request to queue within operation timeout");
ProfileEvents::increment(ProfileEvents::ZooKeeperClose);
}

16
src/Common/ZooKeeper/ZooKeeperImpl.h
View File

@ -7,6 +7,7 @@
#include <Common/ThreadPool.h>
#include <Common/ZooKeeper/IKeeper.h>
#include <Common/ZooKeeper/ZooKeeperCommon.h>
#include <Common/ZooKeeper/ZooKeeperArgs.h>
#include <Coordination/KeeperConstants.h>
#include <IO/ReadBuffer.h>
@ -27,6 +28,7 @@
#include <cstdint>
#include <optional>
#include <functional>
#include <random>
/** ZooKeeper C++ library, a replacement for libzookeeper.
@ -111,12 +113,7 @@ public:
*/
ZooKeeper(
const Nodes & nodes,
const String & root_path,
const String & auth_scheme,
const String & auth_data,
Poco::Timespan session_timeout_,
Poco::Timespan connection_timeout,
Poco::Timespan operation_timeout_,
const zkutil::ZooKeeperArgs & args_,
std::shared_ptr<ZooKeeperLog> zk_log_);
~ZooKeeper() override;
@ -201,11 +198,12 @@ public:
void setZooKeeperLog(std::shared_ptr<DB::ZooKeeperLog> zk_log_);
private:
String root_path;
ACLs default_acls;
Poco::Timespan session_timeout;
Poco::Timespan operation_timeout;
zkutil::ZooKeeperArgs args;
std::optional<std::bernoulli_distribution> send_inject_fault;
std::optional<std::bernoulli_distribution> recv_inject_fault;
Poco::Net::StreamSocket socket;
/// To avoid excessive getpeername(2) calls.

2
src/Common/ZooKeeper/examples/zkutil_test_async.cpp
View File

@ -5,7 +5,7 @@
int main(int argc, char ** argv)
try
{
zkutil::ZooKeeper zookeeper{"localhost:2181"};
zkutil::ZooKeeper zookeeper{zkutil::ZooKeeperArgs("localhost:2181")};
auto nodes = zookeeper.getChildren("/tmp");

2
src/Common/ZooKeeper/examples/zkutil_test_commands.cpp
View File

@ -16,7 +16,7 @@ try
return 1;
}
ZooKeeper zk(argv[1], "", 5000);
ZooKeeper zk{zkutil::ZooKeeperArgs(argv[1])};
std::cout << "create path" << std::endl;
zk.create("/test", "old", zkutil::CreateMode::Persistent);

3
src/Common/ZooKeeper/examples/zkutil_test_commands_new_lib.cpp
View File

@ -40,7 +40,8 @@ try
}
ZooKeeper zk(nodes, {}, {}, {}, {5, 0}, {0, 50000}, {0, 50000}, nullptr);
zkutil::ZooKeeperArgs args;
ZooKeeper zk(nodes, args, nullptr);
Poco::Event event(true);

14
src/Common/tests/gtest_merge_configs.cpp
View File

@ -43,11 +43,8 @@ clickhouse:
text_log:
database: system
table: text_log
partition_by:
"@remove": "1"
engine:
- "@replace" : "1"
- "ENGINE MergeTree"
partition_by: {"@remove": "1"}
engine: "ENGINE MergeTree"
flush_interval_milliseconds: 7500
level: debug
)YAML";
@ -112,11 +109,8 @@ clickhouse:
text_log :
database: system
table: text_log
partition_by:
"@remove": "1"
engine:
- "@replace" : "1"
- "ENGINE MergeTree"
partition_by: {"@remove": "1"}
engine: "ENGINE MergeTree"
flush_interval_milliseconds: 7500
level: debug
)YAML";

169
src/Common/tests/gtest_yaml_parser.cpp
View File

@ -13,40 +13,12 @@
using namespace DB;
TEST(Common, YamlParserInvalidFile)
TEST(YamlParser, InvalidFile)
{
ASSERT_THROW(YAMLParser::parse("some-non-existing-file.yaml"), Exception);
}
TEST(Common, YamlParserProcessKeysList)
{
auto yaml_file = getFileWithContents("keys-list.yaml", R"YAML(
operator:
access_management: "1"
networks:
- ip: "10.1.6.168"
- ip: "::1"
- ip: "127.0.0.1"
)YAML");
SCOPE_EXIT({ yaml_file->remove(); });
Poco::AutoPtr<Poco::XML::Document> xml = YAMLParser::parse(yaml_file->path());
auto *p_node = xml->getNodeByPath("/clickhouse");
EXPECT_EQ(xmlNodeAsString(p_node), R"CONFIG(<clickhouse>
<operator>
<access_management>1</access_management>
<networks>
<ip>10.1.6.168</ip>
<ip>::1</ip>
<ip>127.0.0.1</ip>
</networks>
</operator>
</clickhouse>
)CONFIG");
}
TEST(Common, YamlParserProcessValuesList)
TEST(YamlParser, ProcessValuesList)
{
auto yaml_file = getFileWithContents("values-list.yaml", R"YAML(
rules:
@ -75,4 +47,141 @@ rules:
)CONFIG");
}
TEST(YamlParser, ProcessKeysList)
{
auto yaml_file = getFileWithContents("keys-list.yaml", R"YAML(
operator:
access_management: 1
networks:
ip:
- 10.1.6.168
- ::1
- 127.0.0.1
)YAML");
SCOPE_EXIT({ yaml_file->remove(); });
Poco::AutoPtr<Poco::XML::Document> xml = YAMLParser::parse(yaml_file->path());
auto *p_node = xml->getNodeByPath("/clickhouse");
EXPECT_EQ(xmlNodeAsString(p_node), R"CONFIG(<clickhouse>
<operator>
<access_management>1</access_management>
<networks>
<ip>10.1.6.168</ip>
<ip>::1</ip>
<ip>127.0.0.1</ip>
</networks>
</operator>
</clickhouse>
)CONFIG");
}
TEST(YamlParser, ProcessListAttributes)
{
auto yaml_file = getFileWithContents("list_attributes.yaml", R"YAML(
seq:
- "@attr1": x
- k1: val1
k2: val2
"@attr2": y
- k3: val3
"@attr3": z
)YAML");
SCOPE_EXIT({ yaml_file->remove(); });
Poco::AutoPtr<Poco::XML::Document> xml = YAMLParser::parse(yaml_file->path());
auto *p_node = xml->getNodeByPath("/clickhouse");
EXPECT_EQ(xmlNodeAsString(p_node), R"CONFIG(<clickhouse>
<seq attr1="x"></seq>
<seq attr2="y">
<k1>val1</k1>
<k2>val2</k2>
</seq>
<seq attr3="z">
<k3>val3</k3>
</seq>
</clickhouse>
)CONFIG");
}
TEST(YamlParser, ProcessMapAttributes)
{
auto yaml_file = getFileWithContents("map_attributes.yaml", R"YAML(
map:
"@attr1": x
k1: val1
k2: val2
"@attr2": y
k3: val3
"@attr3": z
)YAML");
SCOPE_EXIT({ yaml_file->remove(); });
Poco::AutoPtr<Poco::XML::Document> xml = YAMLParser::parse(yaml_file->path());
auto *p_node = xml->getNodeByPath("/clickhouse");
EXPECT_EQ(xmlNodeAsString(p_node), R"CONFIG(<clickhouse>
<map attr1="x" attr2="y" attr3="z">
<k1>val1</k1>
<k2>val2</k2>
<k3>val3</k3>
</map>
</clickhouse>
)CONFIG");
}
TEST(YamlParser, ClusterDef)
{
auto yaml_file = getFileWithContents("cluster_def.yaml", R"YAML(
test_cluster:
shard:
- internal_replication: false
replica:
- host: 127.0.0.1
port: 9000
- host: 127.0.0.2
port: 9000
- internal_replication: true
replica:
- host: 127.0.0.3
port: 9000
- host: 127.0.0.4
port: 9000
)YAML");
SCOPE_EXIT({ yaml_file->remove(); });
Poco::AutoPtr<Poco::XML::Document> xml = YAMLParser::parse(yaml_file->path());
auto *p_node = xml->getNodeByPath("/clickhouse");
EXPECT_EQ(xmlNodeAsString(p_node), R"CONFIG(<clickhouse>
<test_cluster>
<shard>
<internal_replication>false</internal_replication>
<replica>
<host>127.0.0.1</host>
<port>9000</port>
</replica>
<replica>
<host>127.0.0.2</host>
<port>9000</port>
</replica>
</shard>
<shard>
<internal_replication>true</internal_replication>
<replica>
<host>127.0.0.3</host>
<port>9000</port>
</replica>
<replica>
<host>127.0.0.4</host>
<port>9000</port>
</replica>
</shard>
</test_cluster>
</clickhouse>
)CONFIG");
}
#endif

28
src/Compression/LZ4_decompress_faster.cpp
View File

@ -478,11 +478,7 @@ template <> void inline copyOverlap<32, true>(UInt8 * op, const UInt8 *& match,
/// See also https://stackoverflow.com/a/30669632
template <size_t copy_amount, bool use_shuffle>
bool NO_INLINE decompressImpl(
const char * const source,
char * const dest,
size_t source_size,
size_t dest_size)
bool NO_INLINE decompressImpl(const char * const source, char * const dest, size_t source_size, size_t dest_size)
{
const UInt8 * ip = reinterpret_cast<const UInt8 *>(source);
UInt8 * op = reinterpret_cast<UInt8 *>(dest);
@ -515,6 +511,18 @@ bool NO_INLINE decompressImpl(
const unsigned token = *ip++;
length = token >> 4;
UInt8 * copy_end;
size_t real_length;
/// It might be true fairly often for well-compressed columns.
/// ATST it may hurt performance in other cases because this condition is hard to predict (especially if the number of zeros is ~50%).
/// In such cases this `if` will significantly increase number of mispredicted instructions. But seems like it results in a
/// noticeable slowdown only for implementations with `copy_amount` > 8. Probably because they use havier instructions.
if constexpr (copy_amount == 8)
if (length == 0)
goto decompress_match;
if (length == 0x0F)
{
if (unlikely(ip + 1 >= input_end))
@ -524,7 +532,7 @@ bool NO_INLINE decompressImpl(
/// Copy literals.
UInt8 * copy_end = op + length;
copy_end = op + length;
/// input: Hello, world
/// ^-ip
@ -541,7 +549,7 @@ bool NO_INLINE decompressImpl(
return false;
// Due to implementation specifics the copy length is always a multiple of copy_amount
size_t real_length = 0;
real_length = 0;
static_assert(copy_amount == 8 || copy_amount == 16 || copy_amount == 32);
if constexpr (copy_amount == 8)
@ -552,9 +560,9 @@ bool NO_INLINE decompressImpl(
real_length = (((length >> 5) + 1) * 32);
if (unlikely(ip + real_length >= input_end + ADDITIONAL_BYTES_AT_END_OF_BUFFER))
return false;
return false;
wildCopy<copy_amount>(op, ip, copy_end); /// Here we can write up to copy_amount - 1 bytes after buffer.
wildCopy<copy_amount>(op, ip, copy_end); /// Here we can write up to copy_amount - 1 bytes after buffer.
if (copy_end == output_end)
return true;
@ -562,6 +570,8 @@ bool NO_INLINE decompressImpl(
ip += length;
op = copy_end;
decompress_match:
if (unlikely(ip + 1 >= input_end))
return false;

19
src/Dictionaries/getDictionaryConfigurationFromAST.cpp
View File

@ -44,15 +44,6 @@ struct AttributeConfiguration
using AttributeNameToConfiguration = std::unordered_map<std::string, AttributeConfiguration>;
/// Get value from field and convert it to string.
/// Also remove quotes from strings.
String getFieldAsString(const Field & field)
{
if (field.getType() == Field::Types::Which::String)
return field.get<String>();
return applyVisitor(FieldVisitorToString(), field);
}
String getAttributeExpression(const ASTDictionaryAttributeDeclaration * dict_attr)
{
if (!dict_attr->expression)
@ -61,7 +52,7 @@ String getAttributeExpression(const ASTDictionaryAttributeDeclaration * dict_att
/// EXPRESSION PROPERTY should be expression or string
String expression_str;
if (const auto * literal = dict_attr->expression->as<ASTLiteral>(); literal && literal->value.getType() == Field::Types::String)
expression_str = getFieldAsString(literal->value);
expression_str = convertFieldToString(literal->value);
else
expression_str = queryToString(dict_attr->expression);
@ -275,7 +266,7 @@ void buildSingleAttribute(
AutoPtr<Element> null_value_element(doc->createElement("null_value"));
String null_value_str;
if (dict_attr->default_value)
null_value_str = getFieldAsString(dict_attr->default_value->as<ASTLiteral>()->value);
null_value_str = convertFieldToString(dict_attr->default_value->as<ASTLiteral>()->value);
AutoPtr<Text> null_value(doc->createTextNode(null_value_str));
null_value_element->appendChild(null_value);
attribute_element->appendChild(null_value_element);
@ -452,7 +443,7 @@ void buildConfigurationFromFunctionWithKeyValueArguments(
}
else if (const auto * literal = pair->second->as<const ASTLiteral>())
{
AutoPtr<Text> value(doc->createTextNode(getFieldAsString(literal->value)));
AutoPtr<Text> value(doc->createTextNode(convertFieldToString(literal->value)));
current_xml_element->appendChild(value);
}
else if (const auto * list = pair->second->as<const ASTExpressionList>())
@ -473,7 +464,7 @@ void buildConfigurationFromFunctionWithKeyValueArguments(
Field value;
result->get(0, value);
AutoPtr<Text> text_value(doc->createTextNode(getFieldAsString(value)));
AutoPtr<Text> text_value(doc->createTextNode(convertFieldToString(value)));
current_xml_element->appendChild(text_value);
}
else
@ -519,7 +510,7 @@ void buildSourceConfiguration(
{
AutoPtr<Element> setting_change_element(doc->createElement(name));
settings_element->appendChild(setting_change_element);
AutoPtr<Text> setting_value(doc->createTextNode(getFieldAsString(value)));
AutoPtr<Text> setting_value(doc->createTextNode(convertFieldToString(value)));
setting_change_element->appendChild(setting_value);
}
}

11
src/Disks/IDisk.h
View File

@ -239,7 +239,16 @@ public:
}
/// For one local path there might be multiple remote paths in case of Log family engines.
using LocalPathWithObjectStoragePaths = std::pair<String, StoredObjects>;
struct LocalPathWithObjectStoragePaths
{
std::string local_path;
std::string common_prefix_for_objects;
StoredObjects objects;
LocalPathWithObjectStoragePaths(
const std::string & local_path_, const std::string & common_prefix_for_objects_, StoredObjects && objects_)
: local_path(local_path_), common_prefix_for_objects(common_prefix_for_objects_), objects(std::move(objects_)) {}
};
virtual void getRemotePathsRecursive(const String &, std::vector<LocalPathWithObjectStoragePaths> &)
{

2
src/Disks/IO/ReadBufferFromAzureBlobStorage.cpp
View File

@ -29,7 +29,7 @@ ReadBufferFromAzureBlobStorage::ReadBufferFromAzureBlobStorage(
size_t max_single_download_retries_,
bool use_external_buffer_,
size_t read_until_position_)
: ReadBufferFromFileBase(read_settings_.remote_fs_buffer_size, nullptr, 0)
: ReadBufferFromFileBase(use_external_buffer_ ? 0 : read_settings_.remote_fs_buffer_size, nullptr, 0)
, blob_container_client(blob_container_client_)
, path(path_)
, max_single_read_retries(max_single_read_retries_)

5
src/Disks/IO/ReadIndirectBufferFromRemoteFS.cpp
View File

@ -1,6 +1,7 @@
#include "ReadIndirectBufferFromRemoteFS.h"
#include <Disks/IO/ReadBufferFromRemoteFSGather.h>
#include <IO/ReadSettings.h>
namespace DB
@ -13,8 +14,8 @@ namespace ErrorCodes
ReadIndirectBufferFromRemoteFS::ReadIndirectBufferFromRemoteFS(
std::shared_ptr<ReadBufferFromRemoteFSGather> impl_)
: ReadBufferFromFileBase(DBMS_DEFAULT_BUFFER_SIZE, nullptr, 0)
std::shared_ptr<ReadBufferFromRemoteFSGather> impl_, const ReadSettings & settings)
: ReadBufferFromFileBase(settings.remote_fs_buffer_size, nullptr, 0)
, impl(impl_)
{
}

3
src/Disks/IO/ReadIndirectBufferFromRemoteFS.h
View File

@ -9,6 +9,7 @@ namespace DB
{
class ReadBufferFromRemoteFSGather;
struct ReadSettings;
/**
* Reads data from S3/HDFS/Web using stored paths in metadata.
@ -18,7 +19,7 @@ class ReadIndirectBufferFromRemoteFS : public ReadBufferFromFileBase
{
public:
explicit ReadIndirectBufferFromRemoteFS(std::shared_ptr<ReadBufferFromRemoteFSGather> impl_);
explicit ReadIndirectBufferFromRemoteFS(std::shared_ptr<ReadBufferFromRemoteFSGather> impl_, const ReadSettings & settings);
off_t seek(off_t offset_, int whence) override;

2
src/Disks/ObjectStorages/AzureBlobStorage/AzureObjectStorage.cpp
View File

@ -112,7 +112,7 @@ std::unique_ptr<ReadBufferFromFileBase> AzureObjectStorage::readObjects( /// NOL
}
else
{
auto buf = std::make_unique<ReadIndirectBufferFromRemoteFS>(std::move(reader_impl));
auto buf = std::make_unique<ReadIndirectBufferFromRemoteFS>(std::move(reader_impl), disk_read_settings);
return std::make_unique<SeekAvoidingReadBuffer>(std::move(buf), settings_ptr->min_bytes_for_seek);
}
}

9
src/Disks/ObjectStorages/DiskObjectStorage.cpp
View File

@ -127,7 +127,7 @@ void DiskObjectStorage::getRemotePathsRecursive(const String & local_path, std::
{
try
{
paths_map.emplace_back(local_path, getStorageObjects(local_path));
paths_map.emplace_back(local_path, metadata_storage->getObjectStorageRootPath(), getStorageObjects(local_path));
}
catch (const Exception & e)
{
@ -253,6 +253,13 @@ void DiskObjectStorage::removeSharedFile(const String & path, bool delete_metada
transaction->commit();
}
void DiskObjectStorage::removeSharedFiles(const RemoveBatchRequest & files, bool keep_all_batch_data, const NameSet & file_names_remove_metadata_only)
{
auto transaction = createObjectStorageTransaction();
transaction->removeSharedFiles(files, keep_all_batch_data, file_names_remove_metadata_only);
transaction->commit();
}
UInt32 DiskObjectStorage::getRefCount(const String & path) const
{
return metadata_storage->getHardlinkCount(path);

2
src/Disks/ObjectStorages/DiskObjectStorage.h
View File

@ -92,6 +92,8 @@ public:
void removeSharedRecursive(const String & path, bool keep_all_batch_data, const NameSet & file_names_remove_metadata_only) override;
void removeSharedFiles(const RemoveBatchRequest & files, bool keep_all_batch_data, const NameSet & file_names_remove_metadata_only) override;
MetadataStoragePtr getMetadataStorage() override { return metadata_storage; }
UInt32 getRefCount(const String & path) const override;

8
src/Disks/ObjectStorages/DiskObjectStorageMetadata.cpp
View File

@ -68,6 +68,14 @@ void DiskObjectStorageMetadata::deserialize(ReadBuffer & buf)
}
}
void DiskObjectStorageMetadata::createFromSingleObject(const std::string & relative_path, size_t bytes_size, size_t ref_count_, bool read_only_)
{
storage_objects.emplace_back(relative_path, bytes_size);
total_size = bytes_size;
ref_count = ref_count_;
read_only = read_only_;
}
void DiskObjectStorageMetadata::deserializeFromString(const std::string & data)
{
ReadBufferFromString buf(data);

1
src/Disks/ObjectStorages/DiskObjectStorageMetadata.h
View File

@ -50,6 +50,7 @@ public:
void deserialize(ReadBuffer & buf);
void deserializeFromString(const std::string & data);
void createFromSingleObject(const std::string & relative_path, size_t bytes_size, size_t ref_count_, bool is_read_only_);
void serialize(WriteBuffer & buf, bool sync) const;
std::string serializeToString() const;

92
src/Disks/ObjectStorages/DiskObjectStorageTransaction.cpp
View File

@ -5,7 +5,6 @@
#include <Common/logger_useful.h>
#include <Common/Exception.h>
namespace DB
{
@ -139,6 +138,87 @@ struct RemoveObjectStorageOperation final : public IDiskObjectStorageOperation
}
};
struct RemoveManyObjectStorageOperation final : public IDiskObjectStorageOperation
{
RemoveBatchRequest remove_paths;
bool keep_all_batch_data;
NameSet file_names_remove_metadata_only;
StoredObjects objects_to_remove;
bool remove_from_cache = false;
RemoveManyObjectStorageOperation(
IObjectStorage & object_storage_,
IMetadataStorage & metadata_storage_,
const RemoveBatchRequest & remove_paths_,
bool keep_all_batch_data_,
const NameSet & file_names_remove_metadata_only_)
: IDiskObjectStorageOperation(object_storage_, metadata_storage_)
, remove_paths(remove_paths_)
, keep_all_batch_data(keep_all_batch_data_)
, file_names_remove_metadata_only(file_names_remove_metadata_only_)
{}
std::string getInfoForLog() const override
{
return fmt::format("RemoveManyObjectStorageOperation (paths size: {}, keep all batch {}, files to keep {})", remove_paths.size(), keep_all_batch_data, fmt::join(file_names_remove_metadata_only, ", "));
}
void execute(MetadataTransactionPtr tx) override
{
for (const auto & [path, if_exists] : remove_paths)
{
if (!metadata_storage.exists(path))
{
if (if_exists)
continue;
throw Exception(ErrorCodes::FILE_DOESNT_EXIST, "Metadata path '{}' doesn't exist", path);
}
if (!metadata_storage.isFile(path))
throw Exception(ErrorCodes::BAD_FILE_TYPE, "Path '{}' is not a regular file", path);
try
{
uint32_t hardlink_count = metadata_storage.getHardlinkCount(path);
auto objects = metadata_storage.getStorageObjects(path);
tx->unlinkMetadata(path);
/// File is really redundant
if (hardlink_count == 0 && !keep_all_batch_data && !file_names_remove_metadata_only.contains(fs::path(path).filename()))
objects_to_remove.insert(objects_to_remove.end(), objects.begin(), objects.end());
}
catch (const Exception & e)
{
/// If it's impossible to read meta - just remove it from FS.
if (e.code() == ErrorCodes::UNKNOWN_FORMAT
|| e.code() == ErrorCodes::ATTEMPT_TO_READ_AFTER_EOF
|| e.code() == ErrorCodes::CANNOT_READ_ALL_DATA
|| e.code() == ErrorCodes::CANNOT_OPEN_FILE)
{
tx->unlinkFile(path);
}
else
throw;
}
}
}
void undo() override
{
}
void finalize() override
{
if (!objects_to_remove.empty())
object_storage.removeObjects(objects_to_remove);
}
};
struct RemoveRecursiveObjectStorageOperation final : public IDiskObjectStorageOperation
{
std::string path;
@ -480,14 +560,8 @@ void DiskObjectStorageTransaction::removeFileIfExists(const std::string & path)
void DiskObjectStorageTransaction::removeSharedFiles(
const RemoveBatchRequest & files, bool keep_all_batch_data, const NameSet & file_names_remove_metadata_only)
{
for (const auto & file : files)
{
bool keep_file = keep_all_batch_data || file_names_remove_metadata_only.contains(fs::path(file.path).filename());
if (file.if_exists)
removeSharedFileIfExists(file.path, keep_file);
else
removeSharedFile(file.path, keep_file);
}
auto operation = std::make_unique<RemoveManyObjectStorageOperation>(object_storage, metadata_storage, files, keep_all_batch_data, file_names_remove_metadata_only);
operations_to_execute.emplace_back(std::move(operation));
}
namespace

5
src/Disks/ObjectStorages/HDFS/HDFSObjectStorage.cpp
View File

@ -70,11 +70,12 @@ std::unique_ptr<ReadBufferFromFileBase> HDFSObjectStorage::readObjects( /// NOLI
auto hdfs_path = path.substr(begin_of_path);
auto hdfs_uri = path.substr(0, begin_of_path);
return std::make_unique<ReadBufferFromHDFS>(hdfs_uri, hdfs_path, config, disk_read_settings);
return std::make_unique<ReadBufferFromHDFS>(
hdfs_uri, hdfs_path, config, disk_read_settings, /* read_until_position */0, /* use_external_buffer */true);
};
auto hdfs_impl = std::make_unique<ReadBufferFromRemoteFSGather>(std::move(read_buffer_creator), objects, disk_read_settings);
auto buf = std::make_unique<ReadIndirectBufferFromRemoteFS>(std::move(hdfs_impl));
auto buf = std::make_unique<ReadIndirectBufferFromRemoteFS>(std::move(hdfs_impl), read_settings);
return std::make_unique<SeekAvoidingReadBuffer>(std::move(buf), settings->min_bytes_for_seek);
}

25
src/Disks/ObjectStorages/S3/S3ObjectStorage.cpp
View File

@ -31,6 +31,7 @@
#include <Common/logger_useful.h>
#include <Common/MultiVersion.h>
namespace DB
{
@ -55,7 +56,7 @@ void throwIfError(const Aws::Utils::Outcome<Result, Error> & response)
if (!response.IsSuccess())
{
const auto & err = response.GetError();
throw Exception(ErrorCodes::S3_ERROR, "{} (Code: {})", err.GetMessage(), static_cast<size_t>(err.GetErrorType()));
throw S3Exception(fmt::format("{} (Code: {})", err.GetMessage(), static_cast<size_t>(err.GetErrorType())), err.GetErrorType());
}
}
@ -69,7 +70,7 @@ void throwIfUnexpectedError(const Aws::Utils::Outcome<Result, Error> & response,
if (!response.IsSuccess() && (!if_exists || !isNotFoundError(response.GetError().GetErrorType())))
{
const auto & err = response.GetError();
throw Exception(ErrorCodes::S3_ERROR, "{} (Code: {})", err.GetMessage(), static_cast<size_t>(err.GetErrorType()));
throw S3Exception(err.GetErrorType(), "{} (Code: {})", err.GetMessage(), static_cast<size_t>(err.GetErrorType()));
}
}
@ -90,7 +91,19 @@ void logIfError(const Aws::Utils::Outcome<Result, Error> & response, std::functi
std::string S3ObjectStorage::generateBlobNameForPath(const std::string & /* path */)
{
return getRandomASCIIString(32);
/// Path to store the new S3 object.
/// Total length is 32 a-z characters for enough randomness.
/// First 3 characters are used as a prefix for
/// https://aws.amazon.com/premiumsupport/knowledge-center/s3-object-key-naming-pattern/
constexpr size_t key_name_total_size = 32;
constexpr size_t key_name_prefix_size = 3;
/// Path to store new S3 object.
return fmt::format("{}/{}",
getRandomASCIIString(key_name_prefix_size),
getRandomASCIIString(key_name_total_size - key_name_prefix_size));
}
Aws::S3::Model::HeadObjectOutcome S3ObjectStorage::requestObjectHeadData(const std::string & bucket_from, const std::string & key) const
@ -157,7 +170,7 @@ std::unique_ptr<ReadBufferFromFileBase> S3ObjectStorage::readObjects( /// NOLINT
}
else
{
auto buf = std::make_unique<ReadIndirectBufferFromRemoteFS>(std::move(s3_impl));
auto buf = std::make_unique<ReadIndirectBufferFromRemoteFS>(std::move(s3_impl), disk_read_settings);
return std::make_unique<SeekAvoidingReadBuffer>(std::move(buf), settings_ptr->min_bytes_for_seek);
}
}
@ -245,6 +258,8 @@ void S3ObjectStorage::removeObjectImpl(const StoredObject & object, bool if_exis
auto outcome = client_ptr->DeleteObject(request);
throwIfUnexpectedError(outcome, if_exists);
LOG_TRACE(log, "Object with path {} was removed from S3", object.absolute_path);
}
void S3ObjectStorage::removeObjectsImpl(const StoredObjects & objects, bool if_exists)
@ -288,6 +303,8 @@ void S3ObjectStorage::removeObjectsImpl(const StoredObjects & objects, bool if_e
auto outcome = client_ptr->DeleteObjects(request);
throwIfUnexpectedError(outcome, if_exists);
LOG_TRACE(log, "Objects with paths [{}] were removed from S3", keys);
}
}
}

2
src/Disks/ObjectStorages/S3/S3ObjectStorage.h
View File

@ -12,6 +12,7 @@
#include <aws/s3/model/ListObjectsV2Result.h>
#include <Storages/StorageS3Settings.h>
#include <Common/MultiVersion.h>
#include <Common/logger_useful.h>
namespace DB
@ -180,6 +181,7 @@ private:
const String version_id;
Poco::Logger * log = &Poco::Logger::get("S3ObjectStorage");
DataSourceDescription data_source_description;
};

3
src/Disks/ObjectStorages/S3/diskSettings.cpp
View File

@ -116,7 +116,8 @@ std::unique_ptr<Aws::S3::S3Client> getClient(const Poco::Util::AbstractConfigura
S3::PocoHTTPClientConfiguration client_configuration = S3::ClientFactory::instance().createClientConfiguration(
config.getString(config_prefix + ".region", ""),
context->getRemoteHostFilter(), context->getGlobalContext()->getSettingsRef().s3_max_redirects,
context->getGlobalContext()->getSettingsRef().enable_s3_requests_logging);
context->getGlobalContext()->getSettingsRef().enable_s3_requests_logging,
/* for_disk_s3 = */ true);
S3::URI uri(Poco::URI(config.getString(config_prefix + ".endpoint")));
if (uri.key.back() != '/')

1
src/Disks/ObjectStorages/StoredObject.h
View File

@ -3,6 +3,7 @@
#include <string>
#include <Disks/ObjectStorages/IObjectStorage_fwd.h>
namespace DB
{

2
src/Disks/ObjectStorages/Web/WebObjectStorage.cpp
View File

@ -188,7 +188,7 @@ std::unique_ptr<ReadBufferFromFileBase> WebObjectStorage::readObject( /// NOLINT
}
else
{
auto buf = std::make_unique<ReadIndirectBufferFromRemoteFS>(std::move(web_impl));
auto buf = std::make_unique<ReadIndirectBufferFromRemoteFS>(std::move(web_impl), read_settings);
return std::make_unique<SeekAvoidingReadBuffer>(std::move(buf), min_bytes_for_seek);
}
}

5
src/Functions/CMakeLists.txt
View File

@ -92,6 +92,11 @@ list (APPEND OBJECT_LIBS $<TARGET_OBJECTS:clickhouse_functions_url>)
add_subdirectory(array)
list (APPEND OBJECT_LIBS $<TARGET_OBJECTS:clickhouse_functions_array>)
if (TARGET ch_contrib::datasketches)
add_subdirectory(UniqTheta)
list (APPEND OBJECT_LIBS $<TARGET_OBJECTS:clickhouse_functions_uniqtheta>)
endif()
add_subdirectory(JSONPath)
list (APPEND PRIVATE_LIBS clickhouse_functions_jsonpath)

4
src/Functions/IFunction.h
View File

@ -171,7 +171,7 @@ public:
*/
virtual bool isSuitableForConstantFolding() const { return true; }
/** If function isSuitableForConstantFolding then, this method will be called during query analyzis
/** If function isSuitableForConstantFolding then, this method will be called during query analysis
* if some arguments are constants. For example logical functions (AndFunction, OrFunction) can
* return they result based on some constant arguments.
* Arguments are passed without modifications, useDefaultImplementationForNulls, useDefaultImplementationForNothing,
@ -394,7 +394,7 @@ private:
using FunctionOverloadResolverPtr = std::shared_ptr<IFunctionOverloadResolver>;
/// Old function interface. Check documentation in IFunction.h.
/// If client do not need statefull properties it can implement this interface.
/// If client do not need stateful properties it can implement this interface.
class IFunction
{
public:

9
src/Functions/UniqTheta/CMakeLists.txt Normal file
View File

@ -0,0 +1,9 @@
include("${ClickHouse_SOURCE_DIR}/cmake/dbms_glob_sources.cmake")
add_library(clickhouse_functions_uniqtheta FunctionsUniqTheta.cpp)
target_link_libraries(clickhouse_functions_uniqtheta PRIVATE dbms)
if (TARGET ch_contrib::datasketches)
target_link_libraries (clickhouse_functions_uniqtheta PRIVATE ch_contrib::datasketches)
endif ()

68
src/Functions/UniqTheta/FunctionsUniqTheta.cpp Normal file
View File

@ -0,0 +1,68 @@
#include <Functions/FunctionFactory.h>
#include "FunctionsUniqTheta.h"
#if USE_DATASKETCHES
namespace DB
{
REGISTER_FUNCTION(UniqTheta)
{
factory.registerFunction<FunctionUniqThetaIntersect>(
{
R"(
Two uniqThetaSketch objects to do intersect calculation(set operation ∩), the result is a new uniqThetaSketch.
A uniqThetaSketch object is to be constructed by aggregation function uniqTheta with -State.
UniqThetaSketch is a data structure storage of approximate values set.
For more information on RoaringBitmap, see: [Theta Sketch Framework](https://datasketches.apache.org/docs/Theta/ThetaSketchFramework.html).
Typical usage:
[example:typical]
)",
Documentation::Examples{
{"typical", "select finalizeAggregation(uniqThetaIntersect(arrayReduce('uniqThetaState',[1,2]), arrayReduce('uniqThetaState',[2,3,4])));"}},
Documentation::Categories{"uniqTheta"}
});
factory.registerFunction<FunctionUniqThetaUnion>(
{
R"(
Two uniqThetaSketch objects to do union calculation(set operation ∪), the result is a new uniqThetaSketch.
A uniqThetaSketch object is to be constructed by aggregation function uniqTheta with -State.
UniqThetaSketch is a data structure storage of approximate values set.
For more information on RoaringBitmap, see: [Theta Sketch Framework](https://datasketches.apache.org/docs/Theta/ThetaSketchFramework.html).
Typical usage:
[example:typical]
)",
Documentation::Examples{
{"typical", "select finalizeAggregation(uniqThetaUnion(arrayReduce('uniqThetaState',[1,2]), arrayReduce('uniqThetaState',[2,3,4])));"}},
Documentation::Categories{"uniqTheta"}
});
factory.registerFunction<FunctionUniqThetaNot>(
{
R"(
Two uniqThetaSketch objects to do a_not_b calculation(set operation ×), the result is a new uniqThetaSketch.
A uniqThetaSketch object is to be constructed by aggregation function uniqTheta with -State.
UniqThetaSketch is a data structure storage of approximate values set.
For more information on RoaringBitmap, see: [Theta Sketch Framework](https://datasketches.apache.org/docs/Theta/ThetaSketchFramework.html).
Typical usage:
[example:typical]
)",
Documentation::Examples{
{"typical", "select finalizeAggregation(uniqThetaNot(arrayReduce('uniqThetaState',[1,2]), arrayReduce('uniqThetaState',[2,3,4])));"}},
Documentation::Categories{"uniqTheta"}
});
}
}
#endif

176
src/Functions/UniqTheta/FunctionsUniqTheta.h Normal file
View File

@ -0,0 +1,176 @@
#pragma once
#include <Common/config.h>
#if USE_DATASKETCHES
#include <AggregateFunctions/AggregateFunctionFactory.h>
#include <Columns/ColumnAggregateFunction.h>
#include <Columns/ColumnArray.h>
#include <Columns/ColumnConst.h>
#include <Columns/ColumnString.h>
#include <Columns/ColumnVector.h>
#include <Columns/ColumnsNumber.h>
#include <DataTypes/DataTypeAggregateFunction.h>
#include <DataTypes/DataTypeArray.h>
#include <DataTypes/DataTypeString.h>
#include <DataTypes/DataTypesNumber.h>
#include <Functions/FunctionHelpers.h>
#include <Functions/IFunction.h>
#include <Interpreters/castColumn.h>
#include <Common/assert_cast.h>
#include <Common/typeid_cast.h>
#include <AggregateFunctions/AggregateFunctionUniq.h>
namespace DB
{
namespace ErrorCodes
{
extern const int ILLEGAL_TYPE_OF_ARGUMENT;
}
struct UniqThetaIntersectImpl
{
static void apply(AggregateFunctionUniqThetaData & sketch_data_1, const AggregateFunctionUniqThetaData & sketch_data_2)
{
sketch_data_1.set.intersect(sketch_data_2.set);
}
};
struct UniqThetaUnionImpl
{
static void apply(AggregateFunctionUniqThetaData & sketch_data_1, const AggregateFunctionUniqThetaData & sketch_data_2)
{
sketch_data_1.set.merge(sketch_data_2.set);
}
};
struct UniqThetaNotImpl
{
static void apply(AggregateFunctionUniqThetaData & sketch_data_1, const AggregateFunctionUniqThetaData & sketch_data_2)
{
sketch_data_1.set.aNotB(sketch_data_2.set);
}
};
template <typename Impl, typename Name>
class FunctionUniqTheta : public IFunction
{
public:
static constexpr auto name = Name::name;
static FunctionPtr create(ContextPtr) { return std::make_shared<FunctionUniqTheta>(); }
String getName() const override { return name; }
bool isVariadic() const override { return false; }
bool isSuitableForShortCircuitArgumentsExecution(const DataTypesWithConstInfo & /*arguments*/) const override { return true; }
size_t getNumberOfArguments() const override { return 2; }
DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
{
const auto * sketch_type0 = typeid_cast<const DataTypeAggregateFunction *>(arguments[0].get());
if (!(sketch_type0 && sketch_type0->getFunctionName() == "uniqTheta"))
throw Exception(
"First argument for function " + getName() + " must be a uniqTheta but it has type " + arguments[0]->getName(),
ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
const auto * sketch_type1 = typeid_cast<const DataTypeAggregateFunction *>(arguments[1].get());
if (!(sketch_type1 && sketch_type1->getFunctionName() == "uniqTheta"))
throw Exception(
"Second argument for function " + getName() + " must be a uniqTheta but it has type " + arguments[1]->getName(),
ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
const DataTypes & arg_data_types0 = sketch_type0->getArgumentsDataTypes();
const DataTypes & arg_data_types1 = sketch_type1->getArgumentsDataTypes();
if (arg_data_types0.size() != arg_data_types1.size())
throw Exception(
"The nested type in uniqThetas must be the same length, but one is " + std::to_string(arg_data_types0.size())
+ ", and the other is " + std::to_string(arg_data_types1.size()),
ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
size_t types_size = arg_data_types0.size();
for (size_t i = 0; i < types_size; ++i)
{
if (!arg_data_types0[i]->equals(*arg_data_types1[i]))
throw Exception(
"The " + std::to_string(i) + "th nested type in uniqThetas must be the same, but one is " + arg_data_types0[i]->getName()
+ ", and the other is " + arg_data_types1[i]->getName(),
ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
}
return arguments[0];
}
bool useDefaultImplementationForConstants() const override { return true; }
ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t input_rows_count) const override
{
const ColumnAggregateFunction * column_ptrs[2];
bool is_column_const[2];
for (size_t i = 0; i < 2; ++i)
{
if (const auto * argument_column_const = typeid_cast<const ColumnConst *>(arguments[i].column.get()))
{
column_ptrs[i] = typeid_cast<const ColumnAggregateFunction *>(argument_column_const->getDataColumnPtr().get());
is_column_const[i] = true;
}
else
{
column_ptrs[i] = typeid_cast<const ColumnAggregateFunction *>(arguments[i].column.get());
is_column_const[i] = false;
}
}
auto col_to = ColumnAggregateFunction::create(column_ptrs[0]->getAggregateFunction());
col_to->reserve(input_rows_count);
const PaddedPODArray<AggregateDataPtr> & container0 = column_ptrs[0]->getData();
const PaddedPODArray<AggregateDataPtr> & container1 = column_ptrs[1]->getData();
for (size_t i = 0; i < input_rows_count; ++i)
{
const AggregateDataPtr data_ptr_0 = is_column_const[0] ? container0[0] : container0[i];
const AggregateDataPtr data_ptr_1 = is_column_const[1] ? container1[0] : container1[i];
col_to->insertFrom(data_ptr_0);
AggregateFunctionUniqThetaData & sketch_data_1 = *reinterpret_cast<AggregateFunctionUniqThetaData *>(col_to->getData()[i]);
const AggregateFunctionUniqThetaData & sketch_data_2
= *reinterpret_cast<const AggregateFunctionUniqThetaData *>(data_ptr_1);
Impl::apply(sketch_data_1, sketch_data_2);
}
return col_to;
}
};
struct NameUniqThetaIntersect
{
static constexpr auto name = "uniqThetaIntersect";
};
struct NameUniqThetaUnion
{
static constexpr auto name = "uniqThetaUnion";
};
struct NameUniqThetaNot
{
static constexpr auto name = "uniqThetaNot";
};
using FunctionUniqThetaIntersect = FunctionUniqTheta<UniqThetaIntersectImpl, NameUniqThetaIntersect>;
using FunctionUniqThetaUnion = FunctionUniqTheta<UniqThetaUnionImpl, NameUniqThetaUnion>;
using FunctionUniqThetaNot = FunctionUniqTheta<UniqThetaNotImpl, NameUniqThetaNot>;
}
#endif

25
src/IO/ReadBufferFromS3.cpp
View File

@ -34,6 +34,7 @@ namespace ErrorCodes
extern const int CANNOT_SEEK_THROUGH_FILE;
extern const int SEEK_POSITION_OUT_OF_BOUND;
extern const int LOGICAL_ERROR;
extern const int CANNOT_ALLOCATE_MEMORY;
}
@ -48,7 +49,7 @@ ReadBufferFromS3::ReadBufferFromS3(
size_t offset_,
size_t read_until_position_,
bool restricted_seek_)
: ReadBufferFromFileBase(settings_.remote_fs_buffer_size, nullptr, 0)
: ReadBufferFromFileBase(use_external_buffer_ ? 0 : settings_.remote_fs_buffer_size, nullptr, 0)
, client_ptr(std::move(client_ptr_))
, bucket(bucket_)
, key(key_)
@ -136,6 +137,23 @@ bool ReadBufferFromS3::nextImpl()
ProfileEvents::increment(ProfileEvents::ReadBufferFromS3Microseconds, watch.elapsedMicroseconds());
ProfileEvents::increment(ProfileEvents::ReadBufferFromS3RequestsErrors, 1);
if (const auto * s3_exception = dynamic_cast<const S3Exception *>(&e))
{
/// It doesn't make sense to retry Access Denied or No Such Key
if (!s3_exception->isRetryableError())
{
tryLogCurrentException(log);
throw;
}
}
/// It doesn't make sense to retry allocator errors
if (e.code() == ErrorCodes::CANNOT_ALLOCATE_MEMORY)
{
tryLogCurrentException(log);
throw;
}
LOG_DEBUG(
log,
"Caught exception while reading S3 object. Bucket: {}, Key: {}, Version: {}, Offset: {}, Attempt: {}, Message: {}",
@ -306,7 +324,10 @@ std::unique_ptr<ReadBuffer> ReadBufferFromS3::initialize()
return std::make_unique<ReadBufferFromIStream>(read_result.GetBody(), buffer_size);
}
else
throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
{
const auto & error = outcome.GetError();
throw S3Exception(error.GetMessage(), error.GetErrorType());
}
}
SeekableReadBufferPtr ReadBufferS3Factory::getReader()

109
src/IO/S3/PocoHTTPClient.cpp
View File

@ -42,6 +42,18 @@ namespace ProfileEvents
extern const Event S3WriteRequestsErrors;
extern const Event S3WriteRequestsThrottling;
extern const Event S3WriteRequestsRedirects;
extern const Event DiskS3ReadMicroseconds;
extern const Event DiskS3ReadRequestsCount;
extern const Event DiskS3ReadRequestsErrors;
extern const Event DiskS3ReadRequestsThrottling;
extern const Event DiskS3ReadRequestsRedirects;
extern const Event DiskS3WriteMicroseconds;
extern const Event DiskS3WriteRequestsCount;
extern const Event DiskS3WriteRequestsErrors;
extern const Event DiskS3WriteRequestsThrottling;
extern const Event DiskS3WriteRequestsRedirects;
}
namespace CurrentMetrics
@ -62,11 +74,13 @@ PocoHTTPClientConfiguration::PocoHTTPClientConfiguration(
const String & force_region_,
const RemoteHostFilter & remote_host_filter_,
unsigned int s3_max_redirects_,
bool enable_s3_requests_logging_)
bool enable_s3_requests_logging_,
bool for_disk_s3_)
: force_region(force_region_)
, remote_host_filter(remote_host_filter_)
, s3_max_redirects(s3_max_redirects_)
, enable_s3_requests_logging(enable_s3_requests_logging_)
, for_disk_s3(for_disk_s3_)
{
}
@ -112,6 +126,7 @@ PocoHTTPClient::PocoHTTPClient(const PocoHTTPClientConfiguration & client_config
, remote_host_filter(client_configuration.remote_host_filter)
, s3_max_redirects(client_configuration.s3_max_redirects)
, enable_s3_requests_logging(client_configuration.enable_s3_requests_logging)
, for_disk_s3(client_configuration.for_disk_s3)
, extra_headers(client_configuration.extra_headers)
{
}
@ -176,6 +191,46 @@ namespace
}
}
PocoHTTPClient::S3MetricKind PocoHTTPClient::getMetricKind(const Aws::Http::HttpRequest & request)
{
switch (request.GetMethod())
{
case Aws::Http::HttpMethod::HTTP_GET:
case Aws::Http::HttpMethod::HTTP_HEAD:
return S3MetricKind::Read;
case Aws::Http::HttpMethod::HTTP_POST:
case Aws::Http::HttpMethod::HTTP_DELETE:
case Aws::Http::HttpMethod::HTTP_PUT:
case Aws::Http::HttpMethod::HTTP_PATCH:
return S3MetricKind::Write;
}
throw Exception("Unsupported request method", ErrorCodes::NOT_IMPLEMENTED);
}
void PocoHTTPClient::addMetric(const Aws::Http::HttpRequest & request, S3MetricType type, ProfileEvents::Count amount) const
{
const ProfileEvents::Event events_map[static_cast<size_t>(S3MetricType::EnumSize)][static_cast<size_t>(S3MetricKind::EnumSize)] = {
{ProfileEvents::S3ReadMicroseconds, ProfileEvents::S3WriteMicroseconds},
{ProfileEvents::S3ReadRequestsCount, ProfileEvents::S3WriteRequestsCount},
{ProfileEvents::S3ReadRequestsErrors, ProfileEvents::S3WriteRequestsErrors},
{ProfileEvents::S3ReadRequestsThrottling, ProfileEvents::S3WriteRequestsThrottling},
{ProfileEvents::S3ReadRequestsRedirects, ProfileEvents::S3WriteRequestsRedirects},
};
const ProfileEvents::Event disk_s3_events_map[static_cast<size_t>(S3MetricType::EnumSize)][static_cast<size_t>(S3MetricKind::EnumSize)] = {
{ProfileEvents::DiskS3ReadMicroseconds, ProfileEvents::DiskS3WriteMicroseconds},
{ProfileEvents::DiskS3ReadRequestsCount, ProfileEvents::DiskS3WriteRequestsCount},
{ProfileEvents::DiskS3ReadRequestsErrors, ProfileEvents::DiskS3WriteRequestsErrors},
{ProfileEvents::DiskS3ReadRequestsThrottling, ProfileEvents::DiskS3WriteRequestsThrottling},
{ProfileEvents::DiskS3ReadRequestsRedirects, ProfileEvents::DiskS3WriteRequestsRedirects},
};
S3MetricKind kind = getMetricKind(request);
ProfileEvents::increment(events_map[static_cast<unsigned int>(type)][static_cast<unsigned int>(kind)], amount);
if (for_disk_s3)
ProfileEvents::increment(disk_s3_events_map[static_cast<unsigned int>(type)][static_cast<unsigned int>(kind)], amount);
}
void PocoHTTPClient::makeRequestInternal(
Aws::Http::HttpRequest & request,
@ -189,45 +244,7 @@ void PocoHTTPClient::makeRequestInternal(
if (enable_s3_requests_logging)
LOG_TEST(log, "Make request to: {}", uri);
enum class S3MetricType
{
Microseconds,
Count,
Errors,
Throttling,
Redirects,
EnumSize,
};
auto select_metric = [&request](S3MetricType type)
{
const ProfileEvents::Event events_map[][2] = {
{ProfileEvents::S3ReadMicroseconds, ProfileEvents::S3WriteMicroseconds},
{ProfileEvents::S3ReadRequestsCount, ProfileEvents::S3WriteRequestsCount},
{ProfileEvents::S3ReadRequestsErrors, ProfileEvents::S3WriteRequestsErrors},
{ProfileEvents::S3ReadRequestsThrottling, ProfileEvents::S3WriteRequestsThrottling},
{ProfileEvents::S3ReadRequestsRedirects, ProfileEvents::S3WriteRequestsRedirects},
};
static_assert((sizeof(events_map) / sizeof(events_map[0])) == static_cast<unsigned int>(S3MetricType::EnumSize));
switch (request.GetMethod())
{
case Aws::Http::HttpMethod::HTTP_GET:
case Aws::Http::HttpMethod::HTTP_HEAD:
return events_map[static_cast<unsigned int>(type)][0]; // Read
case Aws::Http::HttpMethod::HTTP_POST:
case Aws::Http::HttpMethod::HTTP_DELETE:
case Aws::Http::HttpMethod::HTTP_PUT:
case Aws::Http::HttpMethod::HTTP_PATCH:
return events_map[static_cast<unsigned int>(type)][1]; // Write
}
throw Exception("Unsupported request method", ErrorCodes::NOT_IMPLEMENTED);
};
ProfileEvents::increment(select_metric(S3MetricType::Count));
addMetric(request, S3MetricType::Count);
CurrentMetrics::Increment metric_increment{CurrentMetrics::S3Requests};
try
@ -334,7 +351,7 @@ void PocoHTTPClient::makeRequestInternal(
auto & response_body_stream = session->receiveResponse(poco_response);
watch.stop();
ProfileEvents::increment(select_metric(S3MetricType::Microseconds), watch.elapsedMicroseconds());
addMetric(request, S3MetricType::Microseconds, watch.elapsedMicroseconds());
int status_code = static_cast<int>(poco_response.getStatus());
@ -349,7 +366,7 @@ void PocoHTTPClient::makeRequestInternal(
if (enable_s3_requests_logging)
LOG_TEST(log, "Redirecting request to new location: {}", location);
ProfileEvents::increment(select_metric(S3MetricType::Redirects));
addMetric(request, S3MetricType::Redirects);
continue;
}
@ -387,7 +404,7 @@ void PocoHTTPClient::makeRequestInternal(
LOG_WARNING(log, "Response for request contain <Error> tag in body, settings internal server error (500 code)");
response->SetResponseCode(Aws::Http::HttpResponseCode::INTERNAL_SERVER_ERROR);
ProfileEvents::increment(select_metric(S3MetricType::Errors));
addMetric(request, S3MetricType::Errors);
if (error_report)
error_report(request_configuration);
@ -401,11 +418,11 @@ void PocoHTTPClient::makeRequestInternal(
if (status_code == 429 || status_code == 503)
{ // API throttling
ProfileEvents::increment(select_metric(S3MetricType::Throttling));
addMetric(request, S3MetricType::Throttling);
}
else if (status_code >= 300)
{
ProfileEvents::increment(select_metric(S3MetricType::Errors));
addMetric(request, S3MetricType::Errors);
if (status_code >= 500 && error_report)
error_report(request_configuration);
}
@ -423,7 +440,7 @@ void PocoHTTPClient::makeRequestInternal(
response->SetClientErrorType(Aws::Client::CoreErrors::NETWORK_CONNECTION);
response->SetClientErrorMessage(getCurrentExceptionMessage(false));
ProfileEvents::increment(select_metric(S3MetricType::Errors));
addMetric(request, S3MetricType::Errors);
}
}

28
src/IO/S3/PocoHTTPClient.h
View File

@ -44,6 +44,7 @@ struct PocoHTTPClientConfiguration : public Aws::Client::ClientConfiguration
const RemoteHostFilter & remote_host_filter;
unsigned int s3_max_redirects;
bool enable_s3_requests_logging;
bool for_disk_s3;
HeaderCollection extra_headers;
void updateSchemeAndRegion();
@ -55,7 +56,8 @@ private:
const String & force_region_,
const RemoteHostFilter & remote_host_filter_,
unsigned int s3_max_redirects_,
bool enable_s3_requests_logging_
bool enable_s3_requests_logging_,
bool for_disk_s3_
);
/// Constructor of Aws::Client::ClientConfiguration must be called after AWS SDK initialization.
@ -113,18 +115,42 @@ public:
Aws::Utils::RateLimits::RateLimiterInterface * writeLimiter) const override;
private:
void makeRequestInternal(
Aws::Http::HttpRequest & request,
std::shared_ptr<PocoHTTPResponse> & response,
Aws::Utils::RateLimits::RateLimiterInterface * readLimiter,
Aws::Utils::RateLimits::RateLimiterInterface * writeLimiter) const;
enum class S3MetricType
{
Microseconds,
Count,
Errors,
Throttling,
Redirects,
EnumSize,
};
enum class S3MetricKind
{
Read,
Write,
EnumSize,
};
static S3MetricKind getMetricKind(const Aws::Http::HttpRequest & request);
void addMetric(const Aws::Http::HttpRequest & request, S3MetricType type, ProfileEvents::Count amount = 1) const;
std::function<ClientConfigurationPerRequest(const Aws::Http::HttpRequest &)> per_request_configuration;
std::function<void(const ClientConfigurationPerRequest &)> error_report;
ConnectionTimeouts timeouts;
const RemoteHostFilter & remote_host_filter;
unsigned int s3_max_redirects;
bool enable_s3_requests_logging;
bool for_disk_s3;
const HeaderCollection extra_headers;
};

3
src/IO/S3/tests/gtest_aws_s3_client.cpp
View File

@ -87,7 +87,8 @@ TEST(IOTestAwsS3Client, AppendExtraSSECHeaders)
region,
remote_host_filter,
s3_max_redirects,
enable_s3_requests_logging
enable_s3_requests_logging,
/* for_disk_s3 = */ false
);
client_configuration.endpointOverride = uri.endpoint;

29
src/IO/S3Common.cpp
View File

@ -35,6 +35,26 @@
# include <fstream>
namespace DB
{
bool S3Exception::isRetryableError() const
{
/// Looks like these list is quite conservative, add more codes if you wish
static const std::unordered_set<Aws::S3::S3Errors> unretryable_errors = {
Aws::S3::S3Errors::NO_SUCH_KEY,
Aws::S3::S3Errors::ACCESS_DENIED,
Aws::S3::S3Errors::INVALID_ACCESS_KEY_ID,
Aws::S3::S3Errors::INVALID_SIGNATURE,
Aws::S3::S3Errors::NO_SUCH_UPLOAD,
Aws::S3::S3Errors::NO_SUCH_BUCKET,
};
return !unretryable_errors.contains(code);
}
}
namespace
{
@ -543,7 +563,7 @@ public:
/// AWS API tries credentials providers one by one. Some of providers (like ProfileConfigFileAWSCredentialsProvider) can be
/// quite verbose even if nobody configured them. So we use our provider first and only after it use default providers.
{
DB::S3::PocoHTTPClientConfiguration aws_client_configuration = DB::S3::ClientFactory::instance().createClientConfiguration(configuration.region, configuration.remote_host_filter, configuration.s3_max_redirects, configuration.enable_s3_requests_logging);
DB::S3::PocoHTTPClientConfiguration aws_client_configuration = DB::S3::ClientFactory::instance().createClientConfiguration(configuration.region, configuration.remote_host_filter, configuration.s3_max_redirects, configuration.enable_s3_requests_logging, configuration.for_disk_s3);
AddProvider(std::make_shared<AwsAuthSTSAssumeRoleWebIdentityCredentialsProvider>(aws_client_configuration));
}
@ -580,7 +600,7 @@ public:
}
else if (Aws::Utils::StringUtils::ToLower(ec2_metadata_disabled.c_str()) != "true")
{
DB::S3::PocoHTTPClientConfiguration aws_client_configuration = DB::S3::ClientFactory::instance().createClientConfiguration(configuration.region, configuration.remote_host_filter, configuration.s3_max_redirects, configuration.enable_s3_requests_logging);
DB::S3::PocoHTTPClientConfiguration aws_client_configuration = DB::S3::ClientFactory::instance().createClientConfiguration(configuration.region, configuration.remote_host_filter, configuration.s3_max_redirects, configuration.enable_s3_requests_logging, configuration.for_disk_s3);
/// See MakeDefaultHttpResourceClientConfiguration().
/// This is part of EC2 metadata client, but unfortunately it can't be accessed from outside
@ -700,9 +720,10 @@ namespace S3
const String & force_region,
const RemoteHostFilter & remote_host_filter,
unsigned int s3_max_redirects,
bool enable_s3_requests_logging)
bool enable_s3_requests_logging,
bool for_disk_s3)
{
return PocoHTTPClientConfiguration(force_region, remote_host_filter, s3_max_redirects, enable_s3_requests_logging);
return PocoHTTPClientConfiguration(force_region, remote_host_filter, s3_max_redirects, enable_s3_requests_logging, for_disk_s3);
}
URI::URI(const Poco::URI & uri_)

48
src/IO/S3Common.h
View File

@ -7,23 +7,62 @@
#include <base/types.h>
#include <aws/core/Aws.h>
#include <aws/core/client/ClientConfiguration.h>
#include <aws/s3/S3Errors.h>
#include <IO/S3/PocoHTTPClient.h>
#include <Poco/URI.h>
#include <Common/Exception.h>
namespace Aws::S3
{
class S3Client;
}
namespace DB
{
class RemoteHostFilter;
struct HttpHeader;
using HeaderCollection = std::vector<HttpHeader>;
namespace ErrorCodes
{
extern const int S3_ERROR;
}
class RemoteHostFilter;
struct HttpHeader;
using HeaderCollection = std::vector<HttpHeader>;
class S3Exception : public Exception
{
public:
// Format message with fmt::format, like the logging functions.
template <typename... Args>
S3Exception(Aws::S3::S3Errors code_, fmt::format_string<Args...> fmt, Args &&... args)
: Exception(fmt::format(fmt, std::forward<Args>(args)...), ErrorCodes::S3_ERROR)
, code(code_)
{
}
S3Exception(const std::string & msg, Aws::S3::S3Errors code_)
: Exception(msg, ErrorCodes::S3_ERROR)
, code(code_)
{}
Aws::S3::S3Errors getS3ErrorCode() const
{
return code;
}
bool isRetryableError() const;
private:
const Aws::S3::S3Errors code;
};
}
namespace DB::S3
{
class ClientFactory
{
public:
@ -45,7 +84,8 @@ public:
const String & force_region,
const RemoteHostFilter & remote_host_filter,
unsigned int s3_max_redirects,
bool enable_s3_requests_logging);
bool enable_s3_requests_logging,
bool for_disk_s3);
private:
ClientFactory();

102
src/IO/WriteBufferFromS3.cpp
View File

@ -8,6 +8,7 @@
#include <IO/WriteBufferFromS3.h>
#include <IO/WriteHelpers.h>
#include <IO/S3Common.h>
#include <Interpreters/Context.h>
#include <aws/s3/S3Client.h>
@ -40,7 +41,7 @@ namespace ErrorCodes
struct WriteBufferFromS3::UploadPartTask
{
Aws::S3::Model::UploadPartRequest req;
bool is_finised = false;
bool is_finished = false;
std::string tag;
std::exception_ptr exception;
};
@ -48,7 +49,7 @@ struct WriteBufferFromS3::UploadPartTask
struct WriteBufferFromS3::PutObjectTask
{
Aws::S3::Model::PutObjectRequest req;
bool is_finised = false;
bool is_finished = false;
std::exception_ptr exception;
};
@ -64,10 +65,10 @@ WriteBufferFromS3::WriteBufferFromS3(
: BufferWithOwnMemory<WriteBuffer>(buffer_size_, nullptr, 0)
, bucket(bucket_)
, key(key_)
, client_ptr(std::move(client_ptr_))
, upload_part_size(s3_settings_.min_upload_part_size)
, s3_settings(s3_settings_)
, client_ptr(std::move(client_ptr_))
, object_metadata(std::move(object_metadata_))
, upload_part_size(s3_settings_.min_upload_part_size)
, schedule(std::move(schedule_))
, write_settings(write_settings_)
{
@ -173,7 +174,9 @@ void WriteBufferFromS3::finalizeImpl()
auto response = client_ptr->HeadObject(request);
if (!response.IsSuccess())
throw Exception(ErrorCodes::S3_ERROR, "Object {} from bucket {} disappeared immediately after upload, it's a bug in S3 or S3 API.", key, bucket);
throw S3Exception(fmt::format("Object {} from bucket {} disappeared immediately after upload, it's a bug in S3 or S3 API.", key, bucket), response.GetError().GetErrorType());
else
LOG_TRACE(log, "Object {} exists after upload", key);
}
}
@ -197,7 +200,7 @@ void WriteBufferFromS3::createMultipartUpload()
LOG_TRACE(log, "Multipart upload has created. Bucket: {}, Key: {}, Upload id: {}", bucket, key, multipart_upload_id);
}
else
throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
throw S3Exception(outcome.GetError().GetMessage(), outcome.GetError().GetErrorType());
}
void WriteBufferFromS3::writePart()
@ -218,7 +221,7 @@ void WriteBufferFromS3::writePart()
return;
}
if (part_tags.size() == S3_WARN_MAX_PARTS)
if (TSA_SUPPRESS_WARNING_FOR_READ(part_tags).size() == S3_WARN_MAX_PARTS)
{
// Don't throw exception here by ourselves but leave the decision to take by S3 server.
LOG_WARNING(log, "Maximum part number in S3 protocol has reached (too many parts). Server may not accept this whole upload.");
@ -231,6 +234,7 @@ void WriteBufferFromS3::writePart()
int part_number;
{
std::lock_guard lock(bg_tasks_mutex);
task = &upload_object_tasks.emplace_back();
++num_added_bg_tasks;
part_number = num_added_bg_tasks;
@ -240,7 +244,7 @@ void WriteBufferFromS3::writePart()
auto task_finish_notify = [&, task]()
{
std::lock_guard lock(bg_tasks_mutex);
task->is_finised = true;
task->is_finished = true;
++num_finished_bg_tasks;
/// Notification under mutex is important here.
@ -276,9 +280,11 @@ void WriteBufferFromS3::writePart()
else
{
UploadPartTask task;
fillUploadRequest(task.req, part_tags.size() + 1);
auto & tags = TSA_SUPPRESS_WARNING_FOR_WRITE(part_tags); /// Suppress warning because schedule == false.
fillUploadRequest(task.req, tags.size() + 1);
processUploadRequest(task);
part_tags.push_back(task.tag);
tags.push_back(task.tag);
}
}
@ -302,19 +308,22 @@ void WriteBufferFromS3::processUploadRequest(UploadPartTask & task)
if (outcome.IsSuccess())
{
task.tag = outcome.GetResult().GetETag();
std::lock_guard lock(bg_tasks_mutex); /// Protect part_tags from race
LOG_TRACE(log, "Writing part finished. Bucket: {}, Key: {}, Upload_id: {}, Etag: {}, Parts: {}", bucket, key, multipart_upload_id, task.tag, part_tags.size());
}
else
throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
throw S3Exception(outcome.GetError().GetMessage(), outcome.GetError().GetErrorType());
total_parts_uploaded++;
}
void WriteBufferFromS3::completeMultipartUpload()
{
LOG_TRACE(log, "Completing multipart upload. Bucket: {}, Key: {}, Upload_id: {}, Parts: {}", bucket, key, multipart_upload_id, part_tags.size());
const auto & tags = TSA_SUPPRESS_WARNING_FOR_READ(part_tags);
if (part_tags.empty())
LOG_TRACE(log, "Completing multipart upload. Bucket: {}, Key: {}, Upload_id: {}, Parts: {}", bucket, key, multipart_upload_id, tags.size());
if (tags.empty())
throw Exception("Failed to complete multipart upload. No parts have uploaded", ErrorCodes::S3_ERROR);
Aws::S3::Model::CompleteMultipartUploadRequest req;
@ -323,10 +332,10 @@ void WriteBufferFromS3::completeMultipartUpload()
req.SetUploadId(multipart_upload_id);
Aws::S3::Model::CompletedMultipartUpload multipart_upload;
for (size_t i = 0; i < part_tags.size(); ++i)
for (size_t i = 0; i < tags.size(); ++i)
{
Aws::S3::Model::CompletedPart part;
multipart_upload.AddParts(part.WithETag(part_tags[i]).WithPartNumber(i + 1));
multipart_upload.AddParts(part.WithETag(tags[i]).WithPartNumber(i + 1));
}
req.SetMultipartUpload(multipart_upload);
@ -334,12 +343,12 @@ void WriteBufferFromS3::completeMultipartUpload()
auto outcome = client_ptr->CompleteMultipartUpload(req);
if (outcome.IsSuccess())
LOG_TRACE(log, "Multipart upload has completed. Bucket: {}, Key: {}, Upload_id: {}, Parts: {}", bucket, key, multipart_upload_id, part_tags.size());
LOG_TRACE(log, "Multipart upload has completed. Bucket: {}, Key: {}, Upload_id: {}, Parts: {}", bucket, key, multipart_upload_id, tags.size());
else
{
throw Exception(ErrorCodes::S3_ERROR, "{} Tags:{}",
outcome.GetError().GetMessage(),
fmt::join(part_tags.begin(), part_tags.end(), " "));
fmt::join(tags.begin(), tags.end(), " "));
}
}
@ -364,7 +373,7 @@ void WriteBufferFromS3::makeSinglepartUpload()
auto task_notify_finish = [&]()
{
std::lock_guard lock(bg_tasks_mutex);
put_object_task->is_finised = true;
put_object_task->is_finished = true;
/// Notification under mutex is important here.
/// Othervies, WriteBuffer could be destroyed in between
@ -417,37 +426,39 @@ void WriteBufferFromS3::fillPutRequest(Aws::S3::Model::PutObjectRequest & req)
req.SetContentType("binary/octet-stream");
}
void WriteBufferFromS3::processPutRequest(PutObjectTask & task)
void WriteBufferFromS3::processPutRequest(const PutObjectTask & task)
{
auto outcome = client_ptr->PutObject(task.req);
bool with_pool = static_cast<bool>(schedule);
if (outcome.IsSuccess())
LOG_TRACE(log, "Single part upload has completed. Bucket: {}, Key: {}, Object size: {}, WithPool: {}", bucket, key, task.req.GetContentLength(), with_pool);
else
throw Exception(outcome.GetError().GetMessage(), ErrorCodes::S3_ERROR);
throw S3Exception(outcome.GetError().GetMessage(), outcome.GetError().GetErrorType());
}
void WriteBufferFromS3::waitForReadyBackGroundTasks()
{
if (schedule)
{
std::lock_guard lock(bg_tasks_mutex);
std::unique_lock lock(bg_tasks_mutex);
/// Suppress warnings because bg_tasks_mutex is actually hold, but tsa annotations do not understand std::unique_lock
auto & tasks = TSA_SUPPRESS_WARNING_FOR_WRITE(upload_object_tasks);
while (!tasks.empty() && tasks.front().is_finished)
{
while (!upload_object_tasks.empty() && upload_object_tasks.front().is_finised)
auto & task = tasks.front();
auto exception = task.exception;
auto tag = std::move(task.tag);
tasks.pop_front();
if (exception)
{
auto & task = upload_object_tasks.front();
auto exception = task.exception;
auto tag = std::move(task.tag);
upload_object_tasks.pop_front();
if (exception)
{
waitForAllBackGroundTasks();
std::rethrow_exception(exception);
}
part_tags.push_back(tag);
waitForAllBackGroundTasksUnlocked(lock);
std::rethrow_exception(exception);
}
TSA_SUPPRESS_WARNING_FOR_WRITE(part_tags).push_back(tag);
}
}
}
@ -457,22 +468,33 @@ void WriteBufferFromS3::waitForAllBackGroundTasks()
if (schedule)
{
std::unique_lock lock(bg_tasks_mutex);
bg_tasks_condvar.wait(lock, [this]() { return num_added_bg_tasks == num_finished_bg_tasks; });
waitForAllBackGroundTasksUnlocked(lock);
}
}
while (!upload_object_tasks.empty())
void WriteBufferFromS3::waitForAllBackGroundTasksUnlocked(std::unique_lock<std::mutex> & bg_tasks_lock)
{
if (schedule)
{
bg_tasks_condvar.wait(bg_tasks_lock, [this]() {return TSA_SUPPRESS_WARNING_FOR_READ(num_added_bg_tasks) == TSA_SUPPRESS_WARNING_FOR_READ(num_finished_bg_tasks); });
/// Suppress warnings because bg_tasks_mutex is actually hold, but tsa annotations do not understand std::unique_lock
auto & tasks = TSA_SUPPRESS_WARNING_FOR_WRITE(upload_object_tasks);
while (!tasks.empty())
{
auto & task = upload_object_tasks.front();
auto & task = tasks.front();
if (task.exception)
std::rethrow_exception(task.exception);
part_tags.push_back(task.tag);
TSA_SUPPRESS_WARNING_FOR_WRITE(part_tags).push_back(task.tag);
upload_object_tasks.pop_front();
tasks.pop_front();
}
if (put_object_task)
{
bg_tasks_condvar.wait(lock, [this]() { return put_object_task->is_finised; });
bg_tasks_condvar.wait(bg_tasks_lock, [this]() { return put_object_task->is_finished; });
if (put_object_task->exception)
std::rethrow_exception(put_object_task->exception);
}

Some files were not shown because too many files have changed in this diff Show More