Merge branch 'ClickHouse:master' into merge_15765

2024-09-20 16:50:48 +00:00 · 2022-01-25 15:10:40 +03:00 · 2022-01-25 15:10:40 +03:00 · b5be9cbb0d
commit b5be9cbb0d
parent 82b0fc0bd0 fe84167c99
69 changed files with 1373 additions and 290 deletions
--- a/.github/actionlint.yml
+++ b/.github/actionlint.yml
@ -1,8 +1,9 @@
 self-hosted-runner:
  labels:
    - builder
+    - func-tester
+    - func-tester-aarch64
    - fuzzer-unit-tester
    - stress-tester
    - style-checker
-    - func-tester-aarch64
-    - func-tester
+    - style-checker-aarch64
--- a/.github/workflows/backport_branches.yml
+++ b/.github/workflows/backport_branches.yml
@ -10,7 +10,7 @@ on: # yamllint disable-line rule:truthy
      - 'backport/**'
 jobs:
  DockerHubPushAarch64:
-    runs-on: [self-hosted, func-tester-aarch64]
+    runs-on: [self-hosted, style-checker-aarch64]
    steps:
      - name: Clear repository
        run: |
--- a/.github/workflows/docs_check.yml
+++ b/.github/workflows/docs_check.yml
@ -30,7 +30,7 @@ jobs:
          python3 run_check.py
  DockerHubPushAarch64:
    needs: CheckLabels
-    runs-on: [self-hosted, func-tester-aarch64]
+    runs-on: [self-hosted, style-checker-aarch64]
    steps:
      - name: Clear repository
        run: |
--- a/.github/workflows/docs_release.yml
+++ b/.github/workflows/docs_release.yml
@ -20,7 +20,7 @@ on: # yamllint disable-line rule:truthy
  workflow_dispatch:
 jobs:
  DockerHubPushAarch64:
-    runs-on: [self-hosted, func-tester-aarch64]
+    runs-on: [self-hosted, style-checker-aarch64]
    steps:
      - name: Clear repository
        run: |
--- a/.github/workflows/master.yml
+++ b/.github/workflows/master.yml
@ -9,8 +9,20 @@ on: # yamllint disable-line rule:truthy
    branches:
      - 'master'
 jobs:
+  PythonUnitTests:
+    runs-on: [self-hosted, style-checker]
+    steps:
+      - name: Clear repository
+        run: |
+          sudo rm -fr "$GITHUB_WORKSPACE" && mkdir "$GITHUB_WORKSPACE"
+      - name: Check out repository code
+        uses: actions/checkout@v2
+      - name: Python unit tests
+        run: |
+          cd "$GITHUB_WORKSPACE/tests/ci"
+          python3 -m unittest discover -s . -p '*_test.py'
  DockerHubPushAarch64:
-    runs-on: [self-hosted, func-tester-aarch64]
+    runs-on: [self-hosted, style-checker-aarch64]
    steps:
      - name: Clear repository
        run: |
@ -44,7 +56,7 @@ jobs:
          name: changed_images_amd64
          path: ${{ runner.temp }}/docker_images_check/changed_images_amd64.json
  DockerHubPush:
-    needs: [DockerHubPushAmd64, DockerHubPushAarch64]
+    needs: [DockerHubPushAmd64, DockerHubPushAarch64, PythonUnitTests]
    runs-on: [self-hosted, style-checker]
    steps:
      - name: Clear repository
--- a/.github/workflows/pull_request.yml
+++ b/.github/workflows/pull_request.yml
@ -31,9 +31,22 @@ jobs:
        run: |
          cd "$GITHUB_WORKSPACE/tests/ci"
          python3 run_check.py
+  PythonUnitTests:
+    needs: CheckLabels
+    runs-on: [self-hosted, style-checker]
+    steps:
+      - name: Clear repository
+        run: |
+          sudo rm -fr "$GITHUB_WORKSPACE" && mkdir "$GITHUB_WORKSPACE"
+      - name: Check out repository code
+        uses: actions/checkout@v2
+      - name: Python unit tests
+        run: |
+          cd "$GITHUB_WORKSPACE/tests/ci"
+          python3 -m unittest discover -s . -p '*_test.py'
  DockerHubPushAarch64:
    needs: CheckLabels
-    runs-on: [self-hosted, func-tester-aarch64]
+    runs-on: [self-hosted, style-checker-aarch64]
    steps:
      - name: Clear repository
        run: |
@ -68,7 +81,7 @@ jobs:
          name: changed_images_amd64
          path: ${{ runner.temp }}/docker_images_check/changed_images_amd64.json
  DockerHubPush:
-    needs: [DockerHubPushAmd64, DockerHubPushAarch64]
+    needs: [DockerHubPushAmd64, DockerHubPushAarch64, PythonUnitTests]
    runs-on: [self-hosted, style-checker]
    steps:
      - name: Clear repository
--- a/.github/workflows/release_branches.yml
+++ b/.github/workflows/release_branches.yml
@ -13,7 +13,7 @@ on: # yamllint disable-line rule:truthy

 jobs:
  DockerHubPushAarch64:
-    runs-on: [self-hosted, func-tester-aarch64]
+    runs-on: [self-hosted, style-checker-aarch64]
    steps:
      - name: Clear repository
        run: |
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@ -104,9 +104,8 @@ message (STATUS "CMAKE_BUILD_TYPE: ${CMAKE_BUILD_TYPE}")
 string (TOUPPER ${CMAKE_BUILD_TYPE} CMAKE_BUILD_TYPE_UC)

 option(USE_STATIC_LIBRARIES "Disable to use shared libraries" ON)
-option(MAKE_STATIC_LIBRARIES "Disable to make shared libraries" ${USE_STATIC_LIBRARIES})

-if (NOT MAKE_STATIC_LIBRARIES)
+if (NOT USE_STATIC_LIBRARIES)
    # DEVELOPER ONLY.
    # Faster linking if turned on.
    option(SPLIT_SHARED_LIBRARIES "Keep all internal libraries as separate .so files")
@ -115,11 +114,11 @@ if (NOT MAKE_STATIC_LIBRARIES)
        "Make several binaries (clickhouse-server, clickhouse-client etc.) instead of one bundled")
 endif ()

-if (MAKE_STATIC_LIBRARIES AND SPLIT_SHARED_LIBRARIES)
-    message(FATAL_ERROR "Defining SPLIT_SHARED_LIBRARIES=1 without MAKE_STATIC_LIBRARIES=0 has no effect.")
+if (USE_STATIC_LIBRARIES AND SPLIT_SHARED_LIBRARIES)
+    message(FATAL_ERROR "Defining SPLIT_SHARED_LIBRARIES=1 without USE_STATIC_LIBRARIES=0 has no effect.")
 endif()

-if (NOT MAKE_STATIC_LIBRARIES AND SPLIT_SHARED_LIBRARIES)
+if (NOT USE_STATIC_LIBRARIES AND SPLIT_SHARED_LIBRARIES)
    set(BUILD_SHARED_LIBS 1 CACHE INTERNAL "")
 endif ()

@ -201,21 +200,13 @@ endif ()
 option(ENABLE_TESTS "Provide unit_test_dbms target with Google.Test unit tests" ON)
 option(ENABLE_EXAMPLES "Build all example programs in 'examples' subdirectories" OFF)

-if (OS_LINUX AND (ARCH_AMD64 OR ARCH_AARCH64) AND MAKE_STATIC_LIBRARIES AND NOT SPLIT_SHARED_LIBRARIES AND NOT USE_MUSL)
+if (OS_LINUX AND (ARCH_AMD64 OR ARCH_AARCH64) AND USE_STATIC_LIBRARIES AND NOT SPLIT_SHARED_LIBRARIES AND NOT USE_MUSL)
    # Only for Linux, x86_64 or aarch64.
    option(GLIBC_COMPATIBILITY "Enable compatibility with older glibc libraries." ON)
 elseif(GLIBC_COMPATIBILITY)
    message (${RECONFIGURE_MESSAGE_LEVEL} "Glibc compatibility cannot be enabled in current configuration")
 endif ()

-if (GLIBC_COMPATIBILITY)
-    # NOTE: we may also want to check glibc version and add -include only for 2.32+
-    # however this is extra complexity, especially for cross compiling.
-    # And anyway it should not break anything for <2.32.
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -include ${CMAKE_CURRENT_SOURCE_DIR}/base/glibc-compatibility/glibc-compat-2.32.h")
-    set(CMAKE_C_FLAGS   "${CMAKE_C_FLAGS} -include ${CMAKE_CURRENT_SOURCE_DIR}/base/glibc-compatibility/glibc-compat-2.32.h")
-endif()
-
 # Make sure the final executable has symbols exported
 set (CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -rdynamic")

@ -444,7 +435,7 @@ endif ()

 set (CMAKE_POSTFIX_VARIABLE "CMAKE_${CMAKE_BUILD_TYPE_UC}_POSTFIX")

-if (MAKE_STATIC_LIBRARIES)
+if (USE_STATIC_LIBRARIES)
    set (CMAKE_POSITION_INDEPENDENT_CODE OFF)
    if (OS_LINUX AND NOT ARCH_ARM)
        # Slightly more efficient code can be generated
@ -480,7 +471,6 @@ endif ()
 message (STATUS
    "Building for: ${CMAKE_SYSTEM} ${CMAKE_SYSTEM_PROCESSOR} ${CMAKE_LIBRARY_ARCHITECTURE} ;
    USE_STATIC_LIBRARIES=${USE_STATIC_LIBRARIES}
-    MAKE_STATIC_LIBRARIES=${MAKE_STATIC_LIBRARIES}
    SPLIT_SHARED=${SPLIT_SHARED_LIBRARIES}
    CCACHE=${CCACHE_FOUND} ${CCACHE_VERSION}")

@ -528,7 +518,7 @@ macro (add_executable target)
        # - _je_zone_register due to JEMALLOC_PRIVATE_NAMESPACE=je_ under OS X.
        # - but jemalloc-cmake does not run private_namespace.sh
        #   so symbol name should be _zone_register
-        if (ENABLE_JEMALLOC AND MAKE_STATIC_LIBRARIES AND OS_DARWIN)
+        if (ENABLE_JEMALLOC AND USE_STATIC_LIBRARIES AND OS_DARWIN)
            set_property(TARGET ${target} APPEND PROPERTY LINK_OPTIONS -u_zone_register)
        endif()
    endif()
--- a/base/base/CMakeLists.txt
+++ b/base/base/CMakeLists.txt
@ -42,7 +42,7 @@ endif ()

 target_include_directories(common PUBLIC .. "${CMAKE_CURRENT_BINARY_DIR}/..")

-if (OS_DARWIN AND NOT MAKE_STATIC_LIBRARIES)
+if (OS_DARWIN AND NOT USE_STATIC_LIBRARIES)
    target_link_libraries(common PUBLIC -Wl,-U,_inside_main)
 endif()

--- a/base/daemon/CMakeLists.txt
+++ b/base/daemon/CMakeLists.txt
@ -6,7 +6,7 @@ add_library (daemon

 target_include_directories (daemon PUBLIC ..)

-if (OS_DARWIN AND NOT MAKE_STATIC_LIBRARIES)
+if (OS_DARWIN AND NOT USE_STATIC_LIBRARIES)
    target_link_libraries (daemon PUBLIC -Wl,-undefined,dynamic_lookup)
 endif()

--- a/base/glibc-compatibility/CMakeLists.txt
+++ b/base/glibc-compatibility/CMakeLists.txt
@ -37,7 +37,7 @@ if (GLIBC_COMPATIBILITY)

    target_include_directories(glibc-compatibility PRIVATE libcxxabi ${musl_arch_include_dir})

-    if (NOT USE_STATIC_LIBRARIES AND NOT MAKE_STATIC_LIBRARIES)
+    if (NOT USE_STATIC_LIBRARIES AND NOT USE_STATIC_LIBRARIES)
        target_compile_options(glibc-compatibility PRIVATE -fPIC)
    endif ()

--- a/base/glibc-compatibility/glibc-compat-2.32.h
+++ b/base/glibc-compatibility/glibc-compat-2.32.h
@ -1,50 +0,0 @@
-/// In glibc 2.32 new version of some symbols had been added [1]:
-///
-///     $ nm -D clickhouse | fgrep -e @GLIBC_2.32
-///                      U pthread_getattr_np@GLIBC_2.32
-///                      U pthread_sigmask@GLIBC_2.32
-///
-///   [1]: https://www.spinics.net/lists/fedora-devel/msg273044.html
-///
-/// Right now ubuntu 20.04 is used as official image for building
-/// ClickHouse, however once it will be switched someone may not be happy
-/// with that fact that he/she cannot use official binaries anymore because
-/// they have glibc < 2.32.
-///
-/// To avoid this dependency, let's force previous version of those
-/// symbols from glibc.
-///
-/// Also note, that the following approach had been tested:
-/// a) -Wl,--wrap -- but it goes into endless recursion whey you try to do
-///    something like this:
-///
-///     int __pthread_getattr_np_compact(pthread_t thread, pthread_attr_t *attr);
-///     GLIBC_COMPAT_SYMBOL(__pthread_getattr_np_compact, pthread_getattr_np)
-///     int __pthread_getattr_np_compact(pthread_t thread, pthread_attr_t *attr);
-///     int __wrap_pthread_getattr_np(pthread_t thread, pthread_attr_t *attr)
-///     {
-///         return __pthread_getattr_np_compact(thread, attr);
-///     }
-///
-///     int __pthread_sigmask_compact(int how, const sigset_t *set, sigset_t *oldset);
-///     GLIBC_COMPAT_SYMBOL(__pthread_sigmask_compact, pthread_sigmask)
-///     int __pthread_sigmask_compact(int how, const sigset_t *set, sigset_t *oldset);
-///     int __wrap_pthread_sigmask(int how, const sigset_t *set, sigset_t *oldset)
-///     {
-///         return __pthread_sigmask_compact(how, set, oldset);
-///     }
-///
-/// b) -Wl,--defsym -- same problems (and you cannot use version of symbol with
-///    version in the expression)
-/// c) this approach -- simply add this file with -include directive.
-
-#if defined(__amd64__)
-#define GLIBC_COMPAT_SYMBOL(func) __asm__(".symver " #func "," #func "@GLIBC_2.2.5");
-#elif defined(__aarch64__)
-#define GLIBC_COMPAT_SYMBOL(func) __asm__(".symver " #func "," #func "@GLIBC_2.17");
-#else
-#error Your platform is not supported.
-#endif
-
-GLIBC_COMPAT_SYMBOL(pthread_sigmask)
-GLIBC_COMPAT_SYMBOL(pthread_getattr_np)
--- a/cmake/sanitize.cmake
+++ b/cmake/sanitize.cmake
@ -23,7 +23,7 @@ if (SANITIZE)
        if (CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
            set (CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} ${ASAN_FLAGS}")
        endif()
-        if (MAKE_STATIC_LIBRARIES AND CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
+        if (USE_STATIC_LIBRARIES AND CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
            set (CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -static-libasan")
        endif ()
        if (COMPILER_GCC)
@ -48,7 +48,7 @@ if (SANITIZE)
        if (CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
            set (CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -fsanitize=memory")
        endif()
-        if (MAKE_STATIC_LIBRARIES AND CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
+        if (USE_STATIC_LIBRARIES AND CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
            set (CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -static-libmsan")
        endif ()

@ -69,7 +69,7 @@ if (SANITIZE)
        if (CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
            set (CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -fsanitize=thread")
        endif()
-        if (MAKE_STATIC_LIBRARIES AND CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
+        if (USE_STATIC_LIBRARIES AND CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
            set (CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -static-libtsan")
        endif ()
        if (COMPILER_GCC)
@ -101,7 +101,7 @@ if (SANITIZE)
        if (CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
            set (CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -fsanitize=undefined")
        endif()
-        if (MAKE_STATIC_LIBRARIES AND CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
+        if (USE_STATIC_LIBRARIES AND CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
            set (CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -static-libubsan")
        endif ()
        if (COMPILER_GCC)
--- a/contrib/arrow-cmake/CMakeLists.txt
+++ b/contrib/arrow-cmake/CMakeLists.txt
@ -29,7 +29,7 @@ if (OS_FREEBSD)
    message (FATAL_ERROR "Using internal parquet library on FreeBSD is not supported")
 endif()

-if(MAKE_STATIC_LIBRARIES)
+if(USE_STATIC_LIBRARIES)
    set(FLATBUFFERS_LIBRARY flatbuffers)
 else()
    set(FLATBUFFERS_LIBRARY flatbuffers_shared)
@ -84,7 +84,7 @@ set(FLATBUFFERS_BINARY_DIR "${ClickHouse_BINARY_DIR}/contrib/flatbuffers")
 set(FLATBUFFERS_INCLUDE_DIR "${FLATBUFFERS_SRC_DIR}/include")

 # set flatbuffers CMake options
-if (MAKE_STATIC_LIBRARIES)
+if (USE_STATIC_LIBRARIES)
    set(FLATBUFFERS_BUILD_FLATLIB ON CACHE BOOL "Enable the build of the flatbuffers library")
    set(FLATBUFFERS_BUILD_SHAREDLIB OFF CACHE BOOL "Disable the build of the flatbuffers shared library")
 else ()
--- a/contrib/grpc-cmake/CMakeLists.txt
+++ b/contrib/grpc-cmake/CMakeLists.txt
@ -46,7 +46,7 @@ set(_gRPC_SSL_LIBRARIES OpenSSL::Crypto OpenSSL::SSL)
 set(gRPC_ABSL_PROVIDER "clickhouse" CACHE STRING "" FORCE)

 # Choose to build static or shared library for c-ares.
-if (MAKE_STATIC_LIBRARIES)
+if (USE_STATIC_LIBRARIES)
  set(CARES_STATIC ON CACHE BOOL "" FORCE)
  set(CARES_SHARED OFF CACHE BOOL "" FORCE)
 else ()
--- a/contrib/libuv-cmake/CMakeLists.txt
+++ b/contrib/libuv-cmake/CMakeLists.txt
@ -133,7 +133,7 @@ add_library(ch_contrib::uv ALIAS _uv)
 target_compile_definitions(_uv PRIVATE ${uv_defines})
 target_include_directories(_uv SYSTEM PUBLIC ${SOURCE_DIR}/include PRIVATE ${SOURCE_DIR}/src)
 target_link_libraries(_uv ${uv_libraries})
-if (NOT MAKE_STATIC_LIBRARIES)
+if (NOT USE_STATIC_LIBRARIES)
  target_compile_definitions(_uv
                             INTERFACE USING_UV_SHARED=1
                             PRIVATE BUILDING_UV_SHARED=1)
--- a/contrib/lz4
+++ b/contrib/lz4
@ -1 +1 @@
-Subproject commit f39b79fb02962a1cd880bbdecb6dffba4f754a11
+Subproject commit 4c9431e9af596af0556e5da0ae99305bafb2b10b
--- a/docs/en/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-structure.md
+++ b/docs/en/sql-reference/dictionaries/external-dictionaries/external-dicts-dict-structure.md
@ -159,8 +159,7 @@ Configuration fields:
 | Tag                                                  | Description                                                                                                                                                                                                                                                                                                                                     | Required |
 |------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|----------|
 | `name`                                               | Column name.                                                                                                                                                                                                                                                                                                                                    | Yes      |
-| `type`                                               | ClickHouse data type: [UInt8](../../../sql-reference/data-types/int-uint.md), [UInt16](../../../sql-reference/data-types/int-uint.md), [UInt32](../../../sql-reference/data-types/int-uint.md), [UInt64](../../../sql-reference/data-types/int-uint.md), [Int8](../../../sql-reference/data-types/int-uint.md), [Int16](../../../sql-reference/data-types/int-uint.md), [Int32](../../../sql-reference/data-types/int-uint.md), [Int64](../../../sql-reference/data-types/int-uint.md), [Float32](../../../sql-reference/data-types/float.md), [Float64](../../../sql-reference/data-types/float.md), [UUID](../../../sql-reference/data-types/uuid.md), [Decimal32](../../../sql-reference/data-types/decimal.md), [Decimal64](../../../sql-reference/data-types/decimal.md), [Decimal128](../../../sql-reference/data-types/decimal.md), [Decimal256](../../../sql-reference/data-types/decimal.md),
-[Date](../../../sql-reference/data-types/date.md), [Date32](../../../sql-reference/data-types/date32.md), [DateTime](../../../sql-reference/data-types/datetime.md), [DateTime64](../../../sql-reference/data-types/datetime64.md), [String](../../../sql-reference/data-types/string.md), [Array](../../../sql-reference/data-types/array.md).<br/>ClickHouse tries to cast value from dictionary to the specified data type. For example, for MySQL, the field might be `TEXT`, `VARCHAR`, or `BLOB` in the MySQL source table, but it can be uploaded as `String` in ClickHouse.<br/>[Nullable](../../../sql-reference/data-types/nullable.md) is currently supported for [Flat](external-dicts-dict-layout.md#flat), [Hashed](external-dicts-dict-layout.md#dicts-external_dicts_dict_layout-hashed), [ComplexKeyHashed](external-dicts-dict-layout.md#complex-key-hashed), [Direct](external-dicts-dict-layout.md#direct), [ComplexKeyDirect](external-dicts-dict-layout.md#complex-key-direct), [RangeHashed](external-dicts-dict-layout.md#range-hashed), [Polygon](external-dicts-dict-polygon.md), [Cache](external-dicts-dict-layout.md#cache), [ComplexKeyCache](external-dicts-dict-layout.md#complex-key-cache), [SSDCache](external-dicts-dict-layout.md#ssd-cache), [SSDComplexKeyCache](external-dicts-dict-layout.md#complex-key-ssd-cache) dictionaries. In [IPTrie](external-dicts-dict-layout.md#ip-trie) dictionaries `Nullable` types are not supported.       | Yes      |
+| `type`                                               | ClickHouse data type: [UInt8](../../../sql-reference/data-types/int-uint.md), [UInt16](../../../sql-reference/data-types/int-uint.md), [UInt32](../../../sql-reference/data-types/int-uint.md), [UInt64](../../../sql-reference/data-types/int-uint.md), [Int8](../../../sql-reference/data-types/int-uint.md), [Int16](../../../sql-reference/data-types/int-uint.md), [Int32](../../../sql-reference/data-types/int-uint.md), [Int64](../../../sql-reference/data-types/int-uint.md), [Float32](../../../sql-reference/data-types/float.md), [Float64](../../../sql-reference/data-types/float.md), [UUID](../../../sql-reference/data-types/uuid.md), [Decimal32](../../../sql-reference/data-types/decimal.md), [Decimal64](../../../sql-reference/data-types/decimal.md), [Decimal128](../../../sql-reference/data-types/decimal.md), [Decimal256](../../../sql-reference/data-types/decimal.md),[Date](../../../sql-reference/data-types/date.md), [Date32](../../../sql-reference/data-types/date32.md), [DateTime](../../../sql-reference/data-types/datetime.md), [DateTime64](../../../sql-reference/data-types/datetime64.md), [String](../../../sql-reference/data-types/string.md), [Array](../../../sql-reference/data-types/array.md).<br/>ClickHouse tries to cast value from dictionary to the specified data type. For example, for MySQL, the field might be `TEXT`, `VARCHAR`, or `BLOB` in the MySQL source table, but it can be uploaded as `String` in ClickHouse.<br/>[Nullable](../../../sql-reference/data-types/nullable.md) is currently supported for [Flat](external-dicts-dict-layout.md#flat), [Hashed](external-dicts-dict-layout.md#dicts-external_dicts_dict_layout-hashed), [ComplexKeyHashed](external-dicts-dict-layout.md#complex-key-hashed), [Direct](external-dicts-dict-layout.md#direct), [ComplexKeyDirect](external-dicts-dict-layout.md#complex-key-direct), [RangeHashed](external-dicts-dict-layout.md#range-hashed), [Polygon](external-dicts-dict-polygon.md), [Cache](external-dicts-dict-layout.md#cache), [ComplexKeyCache](external-dicts-dict-layout.md#complex-key-cache), [SSDCache](external-dicts-dict-layout.md#ssd-cache), [SSDComplexKeyCache](external-dicts-dict-layout.md#complex-key-ssd-cache) dictionaries. In [IPTrie](external-dicts-dict-layout.md#ip-trie) dictionaries `Nullable` types are not supported.       | Yes      |
 | `null_value`                                         | Default value for a non-existing element.<br/>In the example, it is an empty string. [NULL](../../syntax.md#null-literal) value can be used only for the `Nullable` types (see the previous line with types description).                                                                                                                                                                                                                       | Yes      |
 | `expression`                                         | [Expression](../../../sql-reference/syntax.md#syntax-expressions) that ClickHouse executes on the value.<br/>The expression can be a column name in the remote SQL database. Thus, you can use it to create an alias for the remote column.<br/><br/>Default value: no expression.                                                              | No       |
 | <a name="hierarchical-dict-attr"></a> `hierarchical` | If `true`, the attribute contains the value of a parent key for the current key. See [Hierarchical Dictionaries](../../../sql-reference/dictionaries/external-dictionaries/external-dicts-dict-hierarchical.md).<br/><br/>Default value: `false`.                                                                                               | No       |
--- a/docs/en/sql-reference/statements/create/table.md
+++ b/docs/en/sql-reference/statements/create/table.md
@ -252,7 +252,6 @@ CREATE TABLE codec_example
 ENGINE = MergeTree()
 ```

-
 ### Encryption Codecs {#create-query-encryption-codecs}

 These codecs don't actually compress data, but instead encrypt data on disk. These are only available when an encryption key is specified by [encryption](../../../operations/server-configuration-parameters/settings.md#server-settings-encryption) settings. Note that encryption only makes sense at the end of codec pipelines, because encrypted data usually can't be compressed in any meaningful way.
@ -260,6 +259,7 @@ These codecs don't actually compress data, but instead encrypt data on disk. The
 Encryption codecs:

 -   `CODEC('AES-128-GCM-SIV')` — Encrypts data with AES-128 in [RFC 8452](https://tools.ietf.org/html/rfc8452) GCM-SIV mode. 
+
 -   `CODEC('AES-256-GCM-SIV')` — Encrypts data with AES-256 in GCM-SIV mode. 

 These codecs use a fixed nonce and encryption is therefore deterministic. This makes it compatible with deduplicating engines such as [ReplicatedMergeTree](../../../engines/table-engines/mergetree-family/replication.md) but has a weakness: when the same data block is encrypted twice, the resulting ciphertext will be exactly the same so an adversary who can read the disk can see this equivalence (although only the equivalence, without getting its content).
@ -269,7 +269,7 @@ These codecs use a fixed nonce and encryption is therefore deterministic. This m

 !!! attention "Attention"
    If you perform a SELECT query mentioning a specific value in an encrypted column (such as in its WHERE clause), the value may appear in [system.query_log](../../../operations/system-tables/query_log.md). You may want to disable the logging.
-    
+
 **Example**

 ```sql
--- a/docs/ru/operations/server-configuration-parameters/settings.md
+++ b/docs/ru/operations/server-configuration-parameters/settings.md
@ -105,7 +105,7 @@ ClickHouse проверяет условия для `min_part_size` и `min_part
 ```xml
 <encryption_codecs>
    <aes_128_gcm_siv>
-        <key_hex from_env="KEY"></key_hex>
+        <key_hex from_env="ENVVAR"></key_hex>
    </aes_128_gcm_siv>
 </encryption_codecs>
 ```
@ -118,7 +118,7 @@ ClickHouse проверяет условия для `min_part_size` и `min_part
 <encryption_codecs>
    <aes_128_gcm_siv>
        <key_hex id="0">00112233445566778899aabbccddeeff</key_hex>
-        <key_hex id="1" from_env=".."></key_hex>
+        <key_hex id="1" from_env="ENVVAR"></key_hex>
        <current_key_id>1</current_key_id>
    </aes_128_gcm_siv>
 </encryption_codecs>
--- a/docs/ru/sql-reference/statements/create/table.md
+++ b/docs/ru/sql-reference/statements/create/table.md
@ -246,6 +246,46 @@ CREATE TABLE codec_example
 ENGINE = MergeTree()
 ```

+### Кодеки шифрования {#create-query-encryption-codecs}
+
+Эти кодеки не сжимают данные, вместо этого они зашифровывают данные на диске. Воспользоваться кодеками можно, только когда ключ шифрования задан параметрами [шифрования](../../../operations/server-configuration-parameters/settings.md#server-settings-encryption). Обратите внимание: ставить кодеки шифрования имеет смысл в самый конец цепочки кодеков, потому что зашифрованные данные, как правило, нельзя сжать релевантным образом.
+
+Кодеки шифрования:
+
+-   `CODEC('AES-128-GCM-SIV')` — Зашифровывает данные с помощью AES-128 в режиме [RFC 8452](https://tools.ietf.org/html/rfc8452) GCM-SIV. 
+-   `CODEC('AES-256-GCM-SIV')` — Зашифровывает данные с помощью AES-256 в режиме GCM-SIV. 
+
+Эти кодеки используют фиксированный одноразовый ключ шифрования. Таким образом, это детерминированное шифрование. Оно совместимо с поддерживающими дедупликацию движками, в частности, [ReplicatedMergeTree](../../../engines/table-engines/mergetree-family/replication.md). Однако у шифрования имеется недостаток: если дважды зашифровать один и тот же блок данных, текст на выходе получится одинаковым, и злоумышленник, у которого есть доступ к диску, заметит эту эквивалентность (при этом доступа к содержимому он не получит).
+
+!!! attention "Внимание"
+    Большинство движков, включая семейство `MergeTree`, создают на диске индексные файлы, не применяя кодеки. А значит, в том случае, если зашифрованный столбец индексирован, на диске отобразится незашифрованный текст.
+
+!!! attention "Внимание"
+    Если вы выполняете запрос SELECT с упоминанием конкретного значения в зашифрованном столбце (например, при использовании секции WHERE), это значение может появиться в [system.query_log](../../../operations/system-tables/query_log.md). Рекомендуем отключить логирование.
+
+**Пример**
+
+```sql
+CREATE TABLE mytable 
+(
+    x String Codec(AES_128_GCM_SIV)
+) 
+ENGINE = MergeTree ORDER BY x;
+```
+
+!!!note "Замечание"
+    Если необходимо применить сжатие, это нужно явно прописать в запросе. Без этого будет выполнено только шифрование данных.
+
+**Пример**
+
+```sql
+CREATE TABLE mytable 
+(
+    x String Codec(Delta, LZ4, AES_128_GCM_SIV)
+) 
+ENGINE = MergeTree ORDER BY x;
+```
+
 ## Временные таблицы {#temporary-tables}

 ClickHouse поддерживает временные таблицы со следующими характеристиками:
--- a/docs/zh/faq/general/columnar-database.md
+++ b/docs/zh/faq/general/columnar-database.md
@ -1 +0,0 @@
-../../../en/faq/general/columnar-database.md
--- a/docs/zh/faq/general/columnar-database.md
+++ b/docs/zh/faq/general/columnar-database.md
@ -0,0 +1,25 @@
+---
+title: 什么是列存储数据库?
+toc_hidden: true
+toc_priority: 101
+---
+
+# 什么是列存储数据库? {#what-is-a-columnar-database}
+
+列存储数据库独立存储每个列的数据。这只允许从磁盘读取任何给定查询中使用的列的数据。其代价是，影响整行的操作会按比例变得更昂贵。列存储数据库的同义词是面向列的数据库管理系统。ClickHouse就是这样一个典型的例子。
+
+列存储数据库的主要优点是:
+
+- 查询只使用许多列其中的少数列。
+— 聚合对大量数据的查询。
+— 按列压缩。
+
+下面是构建报表时传统的面向行系统和柱状数据库之间的区别:
+
+**传统行存储**
+!(传统行存储)(https://clickhouse.com/docs/en/images/row-oriented.gif)
+
+**列存储**
+!(列存储)(https://clickhouse.com/docs/en/images/column-oriented.gif)
+
+列存储数据库是分析应用程序的首选，因为它允许在一个表中有许多列以防万一，但不会在读取查询执行时为未使用的列付出代价。面向列的数据库是为大数据处理而设计的，因为和数据仓库一样，它们通常使用分布式的低成本硬件集群来提高吞吐量。ClickHouse结合了[分布式](../../engines/table-engines/special/distributed.md)和[复制式](../../engines/table-engines/mergetree-family/replication.md)两类表。
--- a/programs/CMakeLists.txt
+++ b/programs/CMakeLists.txt
@ -160,7 +160,7 @@ else()
    message(STATUS "ClickHouse keeper-converter mode: OFF")
 endif()

-if(NOT (MAKE_STATIC_LIBRARIES OR SPLIT_SHARED_LIBRARIES))
+if(NOT (USE_STATIC_LIBRARIES OR SPLIT_SHARED_LIBRARIES))
    set(CLICKHOUSE_ONE_SHARED ON)
 endif()

--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@ -11,7 +11,7 @@ if(COMPILER_PIPE)
 else()
    set(MAX_COMPILER_MEMORY 1500)
 endif()
-if(MAKE_STATIC_LIBRARIES)
+if(USE_STATIC_LIBRARIES)
    set(MAX_LINKER_MEMORY 3500)
 else()
    set(MAX_LINKER_MEMORY 2500)
@ -193,7 +193,7 @@ add_subdirectory(Common/Config)

 set (all_modules)
 macro(add_object_library name common_path)
-    if (MAKE_STATIC_LIBRARIES OR NOT SPLIT_SHARED_LIBRARIES)
+    if (USE_STATIC_LIBRARIES OR NOT SPLIT_SHARED_LIBRARIES)
        add_headers_and_sources(dbms ${common_path})
    else ()
        list (APPEND all_modules ${name})
@ -254,7 +254,7 @@ endif()

 set (DBMS_COMMON_LIBRARIES)

-if (MAKE_STATIC_LIBRARIES OR NOT SPLIT_SHARED_LIBRARIES)
+if (USE_STATIC_LIBRARIES OR NOT SPLIT_SHARED_LIBRARIES)
    add_library (dbms STATIC ${dbms_headers} ${dbms_sources})
    target_link_libraries (dbms PRIVATE ch_contrib::libdivide ${DBMS_COMMON_LIBRARIES})
    if (TARGET ch_contrib::jemalloc)
--- a/src/Common/ErrorCodes.cpp
+++ b/src/Common/ErrorCodes.cpp
@ -609,6 +609,7 @@
    M(638, SNAPPY_UNCOMPRESS_FAILED) \
    M(639, SNAPPY_COMPRESS_FAILED) \
    M(640, NO_HIVEMETASTORE) \
+    M(641, CANNOT_APPEND_TO_FILE) \
    \
    M(999, KEEPER_EXCEPTION) \
    M(1000, POCO_EXCEPTION) \
--- a/src/Coordination/KeeperStateMachine.cpp
+++ b/src/Coordination/KeeperStateMachine.cpp
@ -216,7 +216,7 @@ void KeeperStateMachine::create_snapshot(
                std::lock_guard lock(storage_and_responses_lock);
                LOG_TRACE(log, "Clearing garbage after snapshot");
                /// Turn off "snapshot mode" and clear outdate part of storage state
-                storage->clearGarbageAfterSnapshot(snapshot->snapshot_container_size);
+                storage->clearGarbageAfterSnapshot();
                LOG_TRACE(log, "Cleared garbage after snapshot");
                snapshot.reset();
            }
--- a/src/Coordination/KeeperStorage.h
+++ b/src/Coordination/KeeperStorage.h
@ -177,9 +177,9 @@ public:
    }

    /// Clear outdated data from internal container.
-    void clearGarbageAfterSnapshot(size_t up_to_size)
+    void clearGarbageAfterSnapshot()
    {
-        container.clearOutdatedNodes(up_to_size);
+        container.clearOutdatedNodes();
    }

    /// Get all active sessions
--- a/src/Coordination/SnapshotableHashTable.h
+++ b/src/Coordination/SnapshotableHashTable.h
@ -33,6 +33,7 @@ private:
    List list;
    IndexMap map;
    bool snapshot_mode{false};
+    /// Allows to avoid additional copies in updateValue function
    size_t snapshot_up_to_size = 0;
    ArenaWithFreeLists arena;

@ -226,8 +227,10 @@ public:

        if (snapshot_mode)
        {
+            /// We in snapshot mode but updating some node which is already more
+            /// fresh than snapshot distance. So it will not participate in
+            /// snapshot and we don't need to copy it.
            size_t distance = std::distance(list.begin(), list_itr);
-
            if (distance < snapshot_up_to_size)
            {
                auto elem_copy = *(list_itr);
@ -269,11 +272,11 @@ public:
        return it->getMapped()->value;
    }

-    void clearOutdatedNodes(size_t up_to_size)
+    void clearOutdatedNodes()
    {
        auto start = list.begin();
-        size_t counter = 0;
-        for (auto itr = start; counter < up_to_size; ++counter)
+        auto end = list.end();
+        for (auto itr = start; itr != end;)
        {
            if (!itr->active_in_map)
            {
@ -288,7 +291,6 @@ public:
                itr++;
            }
        }
-
    }

    void clear()
@ -308,6 +310,7 @@ public:

    void disableSnapshotMode()
    {
+
        snapshot_mode = false;
        snapshot_up_to_size = 0;
    }
--- a/src/Coordination/tests/gtest_coordination.cpp
+++ b/src/Coordination/tests/gtest_coordination.cpp
@ -908,7 +908,7 @@ TEST_P(CoordinationTest, SnapshotableHashMapTrySnapshot)
        EXPECT_EQ(itr->active_in_map, i != 3 && i != 2);
        itr = std::next(itr);
    }
-    map_snp.clearOutdatedNodes(map_snp.snapshotSize());
+    map_snp.clearOutdatedNodes();

    EXPECT_EQ(map_snp.snapshotSize(), 4);
    EXPECT_EQ(map_snp.size(), 4);
@ -957,13 +957,13 @@ TEST_P(CoordinationTest, SnapshotableHashMapDataSize)
    hello.updateValue("hello", [](IntNode & value) { value = 2; });
    EXPECT_EQ(hello.getApproximateDataSize(), 18);

-    hello.clearOutdatedNodes(hello.snapshotSize());
+    hello.clearOutdatedNodes();
    EXPECT_EQ(hello.getApproximateDataSize(), 9);

    hello.erase("hello");
    EXPECT_EQ(hello.getApproximateDataSize(), 9);

-    hello.clearOutdatedNodes(hello.snapshotSize());
+    hello.clearOutdatedNodes();
    EXPECT_EQ(hello.getApproximateDataSize(), 0);

    /// Node
@ -990,7 +990,7 @@ TEST_P(CoordinationTest, SnapshotableHashMapDataSize)
    world.updateValue("world", [&](Node & value) { value = n2; });
    EXPECT_EQ(world.getApproximateDataSize(), 196);

-    world.clearOutdatedNodes(world.snapshotSize());
+    world.clearOutdatedNodes();
    EXPECT_EQ(world.getApproximateDataSize(), 98);

    world.erase("world");
@ -1170,7 +1170,7 @@ TEST_P(CoordinationTest, TestStorageSnapshotMode)
    }
    EXPECT_TRUE(fs::exists("./snapshots/snapshot_50.bin" + params.extension));
    EXPECT_EQ(storage.container.size(), 26);
-    storage.clearGarbageAfterSnapshot(storage.container.snapshotSize());
+    storage.clearGarbageAfterSnapshot();
    EXPECT_EQ(storage.container.snapshotSize(), 26);
    for (size_t i = 0; i < 50; ++i)
    {
--- a/src/Core/Settings.h
+++ b/src/Core/Settings.h
@ -75,7 +75,11 @@ class IColumn;
    M(UInt64, s3_max_single_read_retries, 4, "The maximum number of retries during single S3 read.", 0) \
    M(UInt64, s3_max_redirects, 10, "Max number of S3 redirects hops allowed.", 0) \
    M(UInt64, s3_max_connections, 1024, "The maximum number of connections per server.", 0) \
+    M(Bool, s3_truncate_on_insert, false, "Enables or disables truncate before insert in s3 engine tables.", 0) \
+    M(Bool, s3_create_new_file_on_insert, false, "Enables or disables creating a new file on each insert in s3 engine tables", 0) \
    M(UInt64, hdfs_replication, 0, "The actual number of replications can be specified when the hdfs file is created.", 0) \
+    M(Bool, hdfs_truncate_on_insert, false, "Enables or disables truncate before insert in s3 engine tables", 0) \
+    M(Bool, hdfs_create_new_file_on_insert, false, "Enables or disables creating a new file on each insert in hdfs engine tables", 0) \
    M(UInt64, hsts_max_age, 0, "Expired time for hsts. 0 means disable HSTS.", 0) \
    M(Bool, extremes, false, "Calculate minimums and maximums of the result columns. They can be output in JSON-formats.", IMPORTANT) \
    M(Bool, use_uncompressed_cache, false, "Whether to use the cache of uncompressed blocks.", 0) \
@ -490,6 +494,7 @@ class IColumn;
    \
    M(Bool, engine_file_empty_if_not_exists, false, "Allows to select data from a file engine table without file", 0) \
    M(Bool, engine_file_truncate_on_insert, false, "Enables or disables truncate before insert in file engine tables", 0) \
+    M(Bool, engine_file_allow_create_multiple_files, false, "Enables or disables creating a new file on each insert in file engine tables if format has suffix.", 0) \
    M(Bool, allow_experimental_database_replicated, false, "Allow to create databases with Replicated engine", 0) \
    M(UInt64, database_replicated_initial_query_timeout_sec, 300, "How long initial DDL query should wait for Replicated database to precess previous DDL queue entries", 0) \
    M(UInt64, max_distributed_depth, 5, "Maximum distributed query depth", 0) \
--- a/src/DataTypes/Serializations/SerializationNumber.cpp
+++ b/src/DataTypes/Serializations/SerializationNumber.cpp
@ -8,7 +8,6 @@
 #include <Common/assert_cast.h>
 #include <Formats/FormatSettings.h>
 #include <Formats/ProtobufReader.h>
-#include <Formats/ProtobufWriter.h>
 #include <Core/Field.h>

 namespace DB
--- a/src/Formats/FormatFactory.cpp
+++ b/src/Formats/FormatFactory.cpp
@ -394,6 +394,27 @@ void FormatFactory::registerNonTrivialPrefixAndSuffixChecker(const String & name
    target = std::move(non_trivial_prefix_and_suffix_checker);
 }

+void FormatFactory::registerAppendSupportChecker(const String & name, AppendSupportChecker append_support_checker)
+{
+    auto & target = dict[name].append_support_checker;
+    if (target)
+        throw Exception("FormatFactory: Suffix checker " + name + " is already registered", ErrorCodes::LOGICAL_ERROR);
+    target = std::move(append_support_checker);
+}
+
+void FormatFactory::markFormatHasNoAppendSupport(const String & name)
+{
+    registerAppendSupportChecker(name, [](const FormatSettings &){ return false; });
+}
+
+bool FormatFactory::checkIfFormatSupportAppend(const String & name, ContextPtr context, const std::optional<FormatSettings> & format_settings_)
+{
+    auto format_settings = format_settings_ ? *format_settings_ : getFormatSettings(context);
+    auto & append_support_checker = dict[name].append_support_checker;
+    /// By default we consider that format supports append
+    return !append_support_checker || append_support_checker(format_settings);
+}
+
 void FormatFactory::registerOutputFormat(const String & name, OutputCreator output_creator)
 {
    auto & target = dict[name].output_creator;
--- a/src/Formats/FormatFactory.h
+++ b/src/Formats/FormatFactory.h
@ -93,6 +93,10 @@ private:
    /// The checker should return true if parallel parsing should be disabled.
    using NonTrivialPrefixAndSuffixChecker = std::function<bool(ReadBuffer & buf)>;

+    /// Some formats can support append depending on settings.
+    /// The checker should return true if format support append.
+    using AppendSupportChecker = std::function<bool(const FormatSettings & settings)>;
+
    using SchemaReaderCreator = std::function<SchemaReaderPtr(ReadBuffer & in, const FormatSettings & settings, ContextPtr context)>;
    using ExternalSchemaReaderCreator = std::function<ExternalSchemaReaderPtr(const FormatSettings & settings)>;

@ -106,6 +110,7 @@ private:
        bool supports_parallel_formatting{false};
        bool is_column_oriented{false};
        NonTrivialPrefixAndSuffixChecker non_trivial_prefix_and_suffix_checker;
+        AppendSupportChecker append_support_checker;
    };

    using FormatsDictionary = std::unordered_map<String, Creators>;
@ -167,6 +172,14 @@ public:

    void registerNonTrivialPrefixAndSuffixChecker(const String & name, NonTrivialPrefixAndSuffixChecker non_trivial_prefix_and_suffix_checker);

+    void registerAppendSupportChecker(const String & name, AppendSupportChecker append_support_checker);
+
+    /// If format always doesn't support append, you can use this method instead of
+    /// registerAppendSupportChecker with append_support_checker that always returns true.
+    void markFormatHasNoAppendSupport(const String & name);
+
+    bool checkIfFormatSupportAppend(const String & name, ContextPtr context, const std::optional<FormatSettings> & format_settings_ = std::nullopt);
+
    /// Register format by its name.
    void registerInputFormat(const String & name, InputCreator input_creator);
    void registerOutputFormat(const String & name, OutputCreator output_creator);
--- a/src/IO/Lz4DeflatingWriteBuffer.cpp
+++ b/src/IO/Lz4DeflatingWriteBuffer.cpp
@ -26,7 +26,7 @@ Lz4DeflatingWriteBuffer::Lz4DeflatingWriteBuffer(
         0 /* no dictID */,
         LZ4F_noBlockChecksum},
        compression_level, /* compression level; 0 == default */
-        0, /* autoflush */
+        1, /* autoflush */
        0, /* favor decompression speed */
        {0, 0, 0}, /* reserved, must be set to 0 */
    };
@ -125,6 +125,8 @@ void Lz4DeflatingWriteBuffer::nextImpl()
        out->position() = out->buffer().begin();
        throw;
    }
+    out->next();
+    out_capacity = out->buffer().end() - out->position();
 }

 void Lz4DeflatingWriteBuffer::finalizeBefore()
--- a/src/IO/Lz4InflatingReadBuffer.cpp
+++ b/src/IO/Lz4InflatingReadBuffer.cpp
@ -70,6 +70,12 @@ bool Lz4InflatingReadBuffer::nextImpl()
        return !working_buffer.empty();
    }

+    /// It may happen that we didn't get new uncompressed data
+    /// (for example if we read the end of frame). Load new data
+    /// in this case.
+    if (working_buffer.empty())
+        return nextImpl();
+
    return true;
 }
 }
--- a/src/IO/ReadBuffer.h
+++ b/src/IO/ReadBuffer.h
@ -63,7 +63,10 @@ public:
        if (!res)
            working_buffer = Buffer(pos, pos);
        else
+        {
            pos = working_buffer.begin() + nextimpl_working_buffer_offset;
+            assert(position() != working_buffer.end());
+        }
        nextimpl_working_buffer_offset = 0;

        assert(position() <= working_buffer.end());
--- a/src/IO/ReadHelpers.cpp
+++ b/src/IO/ReadHelpers.cpp
@ -4,7 +4,6 @@
 #include <Common/StringUtils/StringUtils.h>
 #include <Common/memcpySmall.h>
 #include <Formats/FormatSettings.h>
-#include <IO/WriteHelpers.h>
 #include <IO/WriteBufferFromString.h>
 #include <IO/BufferWithOwnMemory.h>
 #include <IO/readFloatText.h>
--- a/src/Interpreters/AsynchronousMetrics.cpp
+++ b/src/Interpreters/AsynchronousMetrics.cpp
@ -109,6 +109,23 @@ void AsynchronousMetrics::openSensors()
            else
                break;
        }
+
+        file->rewind();
+        Int64 temperature = 0;
+        try
+        {
+            readText(temperature, *file);
+        }
+        catch (const ErrnoException & e)
+        {
+            LOG_WARNING(
+                &Poco::Logger::get("AsynchronousMetrics"),
+                "Thermal monitor '{}' exists but could not be read, error {}.",
+                thermal_device_index,
+                e.getErrno());
+            continue;
+        }
+
        thermal.emplace_back(std::move(file));
    }
 }
@ -222,6 +239,23 @@ void AsynchronousMetrics::openSensorsChips()
                std::replace(sensor_name.begin(), sensor_name.end(), ' ', '_');
            }

+            file->rewind();
+            Int64 temperature = 0;
+            try
+            {
+                readText(temperature, *file);
+            }
+            catch (const ErrnoException & e)
+            {
+                LOG_WARNING(
+                    &Poco::Logger::get("AsynchronousMetrics"),
+                    "Hardware monitor '{}', sensor '{}' exists but could not be read, error {}.",
+                    hwmon_name,
+                    sensor_name,
+                    e.getErrno());
+                continue;
+            }
+
            hwmon_devices[hwmon_name][sensor_name] = std::move(file);
        }
    }
--- a/src/Processors/Formats/Impl/ArrowBlockOutputFormat.cpp
+++ b/src/Processors/Formats/Impl/ArrowBlockOutputFormat.cpp
@ -93,6 +93,7 @@ void registerOutputFormatArrow(FormatFactory & factory)
        {
            return std::make_shared<ArrowBlockOutputFormat>(buf, sample, false, format_settings);
        });
+    factory.markFormatHasNoAppendSupport("Arrow");

    factory.registerOutputFormat(
        "ArrowStream",
@ -103,6 +104,7 @@ void registerOutputFormatArrow(FormatFactory & factory)
        {
            return std::make_shared<ArrowBlockOutputFormat>(buf, sample, true, format_settings);
        });
+    factory.markFormatHasNoAppendSupport("ArrowStream");
 }

 }
--- a/src/Processors/Formats/Impl/AvroRowOutputFormat.cpp
+++ b/src/Processors/Formats/Impl/AvroRowOutputFormat.cpp
@ -479,6 +479,7 @@ void registerOutputFormatAvro(FormatFactory & factory)
    {
        return std::make_shared<AvroRowOutputFormat>(buf, sample, params, settings);
    });
+    factory.markFormatHasNoAppendSupport("Avro");
 }

 }
--- a/src/Processors/Formats/Impl/CustomSeparatedRowOutputFormat.cpp
+++ b/src/Processors/Formats/Impl/CustomSeparatedRowOutputFormat.cpp
@ -91,6 +91,11 @@ void registerOutputFormatCustomSeparated(FormatFactory & factory)
        });

        factory.markOutputFormatSupportsParallelFormatting(format_name);
+
+        factory.registerAppendSupportChecker(format_name, [](const FormatSettings & settings)
+        {
+            return settings.custom.result_after_delimiter.empty();
+        });
    };

    registerWithNamesAndTypes("CustomSeparated", register_func);
--- a/src/Processors/Formats/Impl/JSONRowOutputFormat.cpp
+++ b/src/Processors/Formats/Impl/JSONRowOutputFormat.cpp
@ -284,6 +284,7 @@ void registerOutputFormatJSON(FormatFactory & factory)
    });

    factory.markOutputFormatSupportsParallelFormatting("JSON");
+    factory.markFormatHasNoAppendSupport("JSON");

    factory.registerOutputFormat("JSONStrings", [](
        WriteBuffer & buf,
@ -295,6 +296,7 @@ void registerOutputFormatJSON(FormatFactory & factory)
    });

    factory.markOutputFormatSupportsParallelFormatting("JSONStrings");
+    factory.markFormatHasNoAppendSupport("JSONStrings");
 }

 }
--- a/src/Processors/Formats/Impl/ORCBlockOutputFormat.cpp
+++ b/src/Processors/Formats/Impl/ORCBlockOutputFormat.cpp
@ -526,6 +526,7 @@ void registerOutputFormatORC(FormatFactory & factory)
    {
        return std::make_shared<ORCBlockOutputFormat>(buf, sample, format_settings);
    });
+    factory.markFormatHasNoAppendSupport("ORC");
 }

 }
--- a/src/Processors/Formats/Impl/ParquetBlockOutputFormat.cpp
+++ b/src/Processors/Formats/Impl/ParquetBlockOutputFormat.cpp
@ -85,6 +85,7 @@ void registerOutputFormatParquet(FormatFactory & factory)
        {
            return std::make_shared<ParquetBlockOutputFormat>(buf, sample, format_settings);
        });
+    factory.markFormatHasNoAppendSupport("Parquet");
 }

 }
--- a/src/Processors/Formats/Impl/TemplateBlockOutputFormat.cpp
+++ b/src/Processors/Formats/Impl/TemplateBlockOutputFormat.cpp
@ -235,5 +235,19 @@ void registerOutputFormatTemplate(FormatFactory & factory)

        return std::make_shared<TemplateBlockOutputFormat>(sample, buf, settings, resultset_format, row_format, settings.template_settings.row_between_delimiter);
    });
+
+    factory.registerAppendSupportChecker("Template", [](const FormatSettings & settings)
+    {
+        if (settings.template_settings.resultset_format.empty())
+            return true;
+        auto resultset_format = ParsedTemplateFormatString(
+            FormatSchemaInfo(settings.template_settings.resultset_format, "Template", false,
+                             settings.schema.is_server, settings.schema.format_schema_path),
+            [&](const String & partName)
+            {
+                return static_cast<size_t>(TemplateBlockOutputFormat::stringToResultsetPart(partName));
+            });
+        return resultset_format.delimiters.empty() || resultset_format.delimiters.back().empty();
+    });
 }
 }
--- a/src/Processors/Formats/Impl/XMLRowOutputFormat.cpp
+++ b/src/Processors/Formats/Impl/XMLRowOutputFormat.cpp
@ -256,6 +256,7 @@ void registerOutputFormatXML(FormatFactory & factory)
    });

    factory.markOutputFormatSupportsParallelFormatting("XML");
+    factory.markFormatHasNoAppendSupport("XML");
 }

 }
--- a/src/Storages/HDFS/StorageHDFS.cpp
+++ b/src/Storages/HDFS/StorageHDFS.cpp
@ -14,9 +14,8 @@
 #include <Processors/Formats/IInputFormat.h>
 #include <Processors/Transforms/AddingDefaultsTransform.h>

-#include <IO/ReadHelpers.h>
+#include <IO/WriteHelpers.h>

-#include <Interpreters/Context.h>
 #include <Interpreters/evaluateConstantExpression.h>
 #include <Interpreters/ExpressionAnalyzer.h>
 #include <Interpreters/TreeRewriter.h>
@ -28,7 +27,6 @@
 #include <Storages/HDFS/WriteBufferFromHDFS.h>
 #include <Storages/PartitionedSink.h>

-
 #include <Formats/ReadSchemaUtils.h>
 #include <Formats/FormatFactory.h>
 #include <Functions/FunctionsConversion.h>
@ -52,7 +50,9 @@ namespace ErrorCodes
 {
    extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
    extern const int ACCESS_DENIED;
+    extern const int DATABASE_ACCESS_DENIED;
    extern const int CANNOT_EXTRACT_TABLE_STRUCTURE;
+    extern const int BAD_ARGUMENTS;
    extern const int LOGICAL_ERROR;
 }
 namespace
@ -139,20 +139,23 @@ StorageHDFS::StorageHDFS(
    ASTPtr partition_by_)
    : IStorage(table_id_)
    , WithContext(context_)
-    , uri(uri_)
+    , uris({uri_})
    , format_name(format_name_)
    , compression_method(compression_method_)
    , distributed_processing(distributed_processing_)
    , partition_by(partition_by_)
 {
-    context_->getRemoteHostFilter().checkURL(Poco::URI(uri));
-    checkHDFSURL(uri);
+    context_->getRemoteHostFilter().checkURL(Poco::URI(uri_));
+    checkHDFSURL(uri_);
+
+    String path = uri_.substr(uri_.find('/', uri_.find("//") + 2));
+    is_path_with_globs = path.find_first_of("*?{") != std::string::npos;

    StorageInMemoryMetadata storage_metadata;

    if (columns_.empty())
    {
-        auto columns = getTableStructureFromData(format_name, uri, compression_method, context_);
+        auto columns = getTableStructureFromData(format_name, uri_, compression_method, context_);
        storage_metadata.setColumns(columns);
    }
    else
@ -217,6 +220,39 @@ private:
    Strings::iterator uris_iter;
 };

+class HDFSSource::URISIterator::Impl
+{
+public:
+    explicit Impl(const std::vector<const String> & uris_, ContextPtr context)
+    {
+        auto path_and_uri = getPathFromUriAndUriWithoutPath(uris_[0]);
+        HDFSBuilderWrapper builder = createHDFSBuilder(path_and_uri.second + "/", context->getGlobalContext()->getConfigRef());
+        HDFSFSPtr fs = createHDFSFS(builder.get());
+        for (const auto & uri : uris_)
+        {
+            path_and_uri = getPathFromUriAndUriWithoutPath(uri);
+            if (!hdfsExists(fs.get(), path_and_uri.first.c_str()))
+                uris.push_back(uri);
+        }
+        uris_iter = uris.begin();
+    }
+
+    String next()
+    {
+        std::lock_guard lock(mutex);
+        if (uris_iter == uris.end())
+            return "";
+        auto key = *uris_iter;
+        ++uris_iter;
+        return key;
+    }
+
+private:
+    std::mutex mutex;
+    Strings uris;
+    Strings::iterator uris_iter;
+};
+
 Block HDFSSource::getHeader(const StorageMetadataPtr & metadata_snapshot, bool need_path_column, bool need_file_column)
 {
    auto header = metadata_snapshot->getSampleBlock();
@ -250,6 +286,15 @@ String HDFSSource::DisclosedGlobIterator::next()
    return pimpl->next();
 }

+HDFSSource::URISIterator::URISIterator(const std::vector<const String> & uris_, ContextPtr context)
+    : pimpl(std::make_shared<HDFSSource::URISIterator::Impl>(uris_, context))
+{
+}
+
+String HDFSSource::URISIterator::next()
+{
+    return pimpl->next();
+}

 HDFSSource::HDFSSource(
    StorageHDFSPtr storage_,
@ -284,9 +329,8 @@ bool HDFSSource::initialize()
    current_path = (*file_iterator)();
    if (current_path.empty())
        return false;
-    const size_t begin_of_path = current_path.find('/', current_path.find("//") + 2);
-    const String path_from_uri = current_path.substr(begin_of_path);
-    const String uri_without_path = current_path.substr(0, begin_of_path);
+
+    const auto [path_from_uri, uri_without_path] = getPathFromUriAndUriWithoutPath(current_path);

    auto compression = chooseCompressionMethod(path_from_uri, storage->compression_method);
    read_buf = wrapReadBufferWithCompressionMethod(std::make_unique<ReadBufferFromHDFS>(uri_without_path, path_from_uri, getContext()->getGlobalContext()->getConfigRef()), compression);
@ -469,15 +513,23 @@ Pipe StorageHDFS::read(
                return callback();
        });
    }
-    else
+    else if (is_path_with_globs)
    {
        /// Iterate through disclosed globs and make a source for each file
-        auto glob_iterator = std::make_shared<HDFSSource::DisclosedGlobIterator>(context_, uri);
+        auto glob_iterator = std::make_shared<HDFSSource::DisclosedGlobIterator>(context_, uris[0]);
        iterator_wrapper = std::make_shared<HDFSSource::IteratorWrapper>([glob_iterator]()
        {
            return glob_iterator->next();
        });
    }
+    else
+    {
+        auto uris_iterator = std::make_shared<HDFSSource::URISIterator>(uris, context_);
+        iterator_wrapper = std::make_shared<HDFSSource::IteratorWrapper>([uris_iterator]()
+        {
+            return uris_iterator->next();
+        });
+    }

    Pipes pipes;
    auto this_ptr = std::static_pointer_cast<StorageHDFS>(shared_from_this());
@ -505,9 +557,11 @@ Pipe StorageHDFS::read(
    return Pipe::unitePipes(std::move(pipes));
 }

-SinkToStoragePtr StorageHDFS::write(const ASTPtr & query, const StorageMetadataPtr & metadata_snapshot, ContextPtr /*context*/)
+SinkToStoragePtr StorageHDFS::write(const ASTPtr & query, const StorageMetadataPtr & metadata_snapshot, ContextPtr context_)
 {
-    bool has_wildcards = uri.find(PartitionedSink::PARTITION_ID_WILDCARD) != String::npos;
+    String current_uri = uris.back();
+
+    bool has_wildcards = current_uri.find(PartitionedSink::PARTITION_ID_WILDCARD) != String::npos;
    const auto * insert_query = dynamic_cast<const ASTInsertQuery *>(query.get());
    auto partition_by_ast = insert_query ? (insert_query->partition_by ? insert_query->partition_by : partition_by) : nullptr;
    bool is_partitioned_implementation = partition_by_ast && has_wildcards;
@ -516,34 +570,70 @@ SinkToStoragePtr StorageHDFS::write(const ASTPtr & query, const StorageMetadataP
    {
        return std::make_shared<PartitionedHDFSSink>(
            partition_by_ast,
-            uri,
+            current_uri,
            format_name,
            metadata_snapshot->getSampleBlock(),
-            getContext(),
-            chooseCompressionMethod(uri, compression_method));
+            context_,
+            chooseCompressionMethod(current_uri, compression_method));
    }
    else
    {
-        return std::make_shared<HDFSSink>(uri,
+        if (is_path_with_globs)
+            throw Exception(ErrorCodes::DATABASE_ACCESS_DENIED, "URI '{}' contains globs, so the table is in readonly mode", uris.back());
+
+        const auto [path_from_uri, uri_without_path] = getPathFromUriAndUriWithoutPath(current_uri);
+
+        HDFSBuilderWrapper builder = createHDFSBuilder(uri_without_path + "/", context_->getGlobalContext()->getConfigRef());
+        HDFSFSPtr fs = createHDFSFS(builder.get());
+
+        bool truncate_on_insert = context_->getSettingsRef().hdfs_truncate_on_insert;
+        if (!truncate_on_insert && !hdfsExists(fs.get(), path_from_uri.c_str()))
+        {
+            if (context_->getSettingsRef().hdfs_create_new_file_on_insert)
+            {
+                auto pos = uris[0].find_first_of('.', uris[0].find_last_of('/'));
+                size_t index = uris.size();
+                String new_uri;
+                do
+                {
+                    new_uri = uris[0].substr(0, pos) + "." + std::to_string(index) + (pos == std::string::npos ? "" : uris[0].substr(pos));
+                    ++index;
+                }
+                while (!hdfsExists(fs.get(), new_uri.c_str()));
+                uris.push_back(new_uri);
+                current_uri = new_uri;
+            }
+            else
+                throw Exception(
+                    ErrorCodes::BAD_ARGUMENTS,
+                    "File with path {} already exists. If you want to overwrite it, enable setting hdfs_truncate_on_insert, "
+                    "if you want to create new file on each insert, enable setting hdfs_create_new_file_on_insert",
+                    path_from_uri);
+        }
+
+        return std::make_shared<HDFSSink>(current_uri,
            format_name,
            metadata_snapshot->getSampleBlock(),
-            getContext(),
-            chooseCompressionMethod(uri, compression_method));
+            context_,
+            chooseCompressionMethod(current_uri, compression_method));
    }
 }

 void StorageHDFS::truncate(const ASTPtr & /* query */, const StorageMetadataPtr &, ContextPtr local_context, TableExclusiveLockHolder &)
 {
-    const size_t begin_of_path = uri.find('/', uri.find("//") + 2);
-    const String path = uri.substr(begin_of_path);
-    const String url = uri.substr(0, begin_of_path);
+    const size_t begin_of_path = uris[0].find('/', uris[0].find("//") + 2);
+    const String url = uris[0].substr(0, begin_of_path);

    HDFSBuilderWrapper builder = createHDFSBuilder(url + "/", local_context->getGlobalContext()->getConfigRef());
    HDFSFSPtr fs = createHDFSFS(builder.get());

-    int ret = hdfsDelete(fs.get(), path.data(), 0);
-    if (ret)
-        throw Exception(ErrorCodes::ACCESS_DENIED, "Unable to truncate hdfs table: {}", std::string(hdfsGetLastError()));
+    for (const auto & uri : uris)
+    {
+        const String path = uri.substr(begin_of_path);
+        int ret = hdfsDelete(fs.get(), path.data(), 0);
+        if (ret)
+            throw Exception(ErrorCodes::ACCESS_DENIED, "Unable to truncate hdfs table: {}", std::string(hdfsGetLastError()));
+    }
 }


--- a/src/Storages/HDFS/StorageHDFS.h
+++ b/src/Storages/HDFS/StorageHDFS.h
@ -31,7 +31,7 @@ public:
        size_t max_block_size,
        unsigned num_streams) override;

-    SinkToStoragePtr write(const ASTPtr & query, const StorageMetadataPtr & /*metadata_snapshot*/, ContextPtr /*context*/) override;
+    SinkToStoragePtr write(const ASTPtr & query, const StorageMetadataPtr & metadata_snapshot, ContextPtr context) override;

    void truncate(
        const ASTPtr & query,
@ -70,11 +70,12 @@ protected:
        ASTPtr partition_by = nullptr);

 private:
-    const String uri;
+    std::vector<const String> uris;
    String format_name;
    String compression_method;
    const bool distributed_processing;
    ASTPtr partition_by;
+    bool is_path_with_globs;

    Poco::Logger * log = &Poco::Logger::get("StorageHDFS");
 };
@ -95,6 +96,17 @@ public:
            std::shared_ptr<Impl> pimpl;
    };

+    class URISIterator
+    {
+        public:
+            URISIterator(const std::vector<const String> & uris_, ContextPtr context);
+            String next();
+        private:
+            class Impl;
+            /// shared_ptr to have copy constructor
+            std::shared_ptr<Impl> pimpl;
+    };
+
    using IteratorWrapper = std::function<String()>;
    using StorageHDFSPtr = std::shared_ptr<StorageHDFS>;

--- a/src/Storages/HDFS/WriteBufferFromHDFS.cpp
+++ b/src/Storages/HDFS/WriteBufferFromHDFS.cpp
@ -15,7 +15,6 @@ namespace ErrorCodes
 extern const int NETWORK_ERROR;
 extern const int CANNOT_OPEN_FILE;
 extern const int CANNOT_FSYNC;
-extern const int BAD_ARGUMENTS;
 }


@ -38,12 +37,6 @@ struct WriteBufferFromHDFS::WriteBufferFromHDFSImpl
        const size_t begin_of_path = hdfs_uri.find('/', hdfs_uri.find("//") + 2);
        const String path = hdfs_uri.substr(begin_of_path);

-        if (path.find_first_of("*?{") != std::string::npos)
-            throw Exception(ErrorCodes::CANNOT_OPEN_FILE, "URI '{}' contains globs, so the table is in readonly mode", hdfs_uri);
-
-        if (!hdfsExists(fs.get(), path.c_str()))
-            throw Exception(ErrorCodes::BAD_ARGUMENTS, "File {} already exists", path);
-
        fout = hdfsOpenFile(fs.get(), path.c_str(), flags, 0, replication_, 0);     /// O_WRONLY meaning create or overwrite i.e., implies O_TRUNCAT here

        if (fout == nullptr)
--- a/src/Storages/StorageFile.cpp
+++ b/src/Storages/StorageFile.cpp
@ -65,6 +65,7 @@ namespace ErrorCodes
    extern const int INCOMPATIBLE_COLUMNS;
    extern const int CANNOT_STAT;
    extern const int LOGICAL_ERROR;
+    extern const int CANNOT_APPEND_TO_FILE;
    extern const int CANNOT_EXTRACT_TABLE_STRUCTURE;
 }

@ -285,6 +286,7 @@ StorageFile::StorageFile(const std::string & table_path_, const std::string & us
 {
    is_db_table = false;
    paths = getPathsList(table_path_, user_files_path, args.getContext(), total_bytes_to_read);
+    is_path_with_globs = paths.size() > 1;
    path_for_partitioned_write = table_path_;
    setStorageMetadata(args);
 }
@ -603,7 +605,7 @@ public:
        int table_fd_,
        bool use_table_fd_,
        std::string base_path_,
-        std::vector<std::string> paths_,
+        std::string path_,
        const CompressionMethod compression_method_,
        const std::optional<FormatSettings> & format_settings_,
        const String format_name_,
@ -615,7 +617,7 @@ public:
        , table_fd(table_fd_)
        , use_table_fd(use_table_fd_)
        , base_path(base_path_)
-        , paths(paths_)
+        , path(path_)
        , compression_method(compression_method_)
        , format_name(format_name_)
        , format_settings(format_settings_)
@ -632,7 +634,7 @@ public:
        int table_fd_,
        bool use_table_fd_,
        std::string base_path_,
-        std::vector<std::string> paths_,
+        const std::string & path_,
        const CompressionMethod compression_method_,
        const std::optional<FormatSettings> & format_settings_,
        const String format_name_,
@ -644,7 +646,7 @@ public:
        , table_fd(table_fd_)
        , use_table_fd(use_table_fd_)
        , base_path(base_path_)
-        , paths(paths_)
+        , path(path_)
        , compression_method(compression_method_)
        , format_name(format_name_)
        , format_settings(format_settings_)
@ -666,10 +668,8 @@ public:
        }
        else
        {
-            if (paths.size() != 1)
-                throw Exception("Table '" + table_name_for_log + "' is in readonly mode because of globs in filepath", ErrorCodes::DATABASE_ACCESS_DENIED);
            flags |= O_WRONLY | O_APPEND | O_CREAT;
-            naked_buffer = std::make_unique<WriteBufferFromFile>(paths[0], DBMS_DEFAULT_BUFFER_SIZE, flags);
+            naked_buffer = std::make_unique<WriteBufferFromFile>(path, DBMS_DEFAULT_BUFFER_SIZE, flags);
        }

        /// In case of formats with prefixes if file is not empty we have already written prefix.
@ -709,7 +709,7 @@ private:
    int table_fd;
    bool use_table_fd;
    std::string base_path;
-    std::vector<std::string> paths;
+    std::string path;
    CompressionMethod compression_method;
    std::string format_name;
    std::optional<FormatSettings> format_settings;
@ -752,7 +752,6 @@ public:
    {
        auto partition_path = PartitionedSink::replaceWildcards(path, partition_id);
        PartitionedSink::validatePartitionKey(partition_path, true);
-        Strings result_paths = {partition_path};
        checkCreationIsAllowed(context, context->getUserFilesPath(), partition_path);
        return std::make_shared<StorageFileSink>(
            metadata_snapshot,
@ -760,7 +759,7 @@ public:
            -1,
            /* use_table_fd */false,
            base_path,
-            result_paths,
+            partition_path,
            compression_method,
            format_settings,
            format_name,
@ -794,7 +793,6 @@ SinkToStoragePtr StorageFile::write(

    int flags = 0;

-    std::string path;
    if (context->getSettingsRef().engine_file_truncate_on_insert)
        flags |= O_TRUNC;

@ -815,7 +813,7 @@ SinkToStoragePtr StorageFile::write(
            std::unique_lock{rwlock, getLockTimeout(context)},
            base_path,
            path_for_partitioned_write,
-            chooseCompressionMethod(path, compression_method),
+            chooseCompressionMethod(path_for_partitioned_write, compression_method),
            format_settings,
            format_name,
            context,
@ -823,10 +821,41 @@ SinkToStoragePtr StorageFile::write(
    }
    else
    {
+        String path;
        if (!paths.empty())
        {
-            path = paths[0];
+            if (is_path_with_globs)
+                throw Exception("Table '" + getStorageID().getNameForLogs() + "' is in readonly mode because of globs in filepath", ErrorCodes::DATABASE_ACCESS_DENIED);
+
+            path = paths.back();
            fs::create_directories(fs::path(path).parent_path());
+
+            if (!context->getSettingsRef().engine_file_truncate_on_insert && !is_path_with_globs
+                && !FormatFactory::instance().checkIfFormatSupportAppend(format_name, context, format_settings) && fs::exists(paths.back())
+                && fs::file_size(paths.back()) != 0)
+            {
+                if (context->getSettingsRef().engine_file_allow_create_multiple_files)
+                {
+                    auto pos = paths[0].find_first_of('.', paths[0].find_last_of('/'));
+                    size_t index = paths.size();
+                    String new_path;
+                    do
+                    {
+                        new_path = paths[0].substr(0, pos) + "." + std::to_string(index) + (pos == std::string::npos ? "" : paths[0].substr(pos));
+                        ++index;
+                    }
+                    while (fs::exists(new_path));
+                    paths.push_back(new_path);
+                    path = new_path;
+                }
+                else
+                    throw Exception(
+                        ErrorCodes::CANNOT_APPEND_TO_FILE,
+                        "Cannot append data in format {} to file, because this format doesn't support appends."
+                        " You can allow to create a new file "
+                        "on each insert by enabling setting engine_file_allow_create_multiple_files",
+                        format_name);
+            }
        }

        return std::make_shared<StorageFileSink>(
@ -836,7 +865,7 @@ SinkToStoragePtr StorageFile::write(
            table_fd,
            use_table_fd,
            base_path,
-            paths,
+            path,
            chooseCompressionMethod(path, compression_method),
            format_settings,
            format_name,
@ -882,7 +911,7 @@ void StorageFile::truncate(
    ContextPtr /* context */,
    TableExclusiveLockHolder &)
 {
-    if (paths.size() != 1)
+    if (is_path_with_globs)
        throw Exception("Can't truncate table '" + getStorageID().getNameForLogs() + "' in readonly mode", ErrorCodes::DATABASE_ACCESS_DENIED);

    if (use_table_fd)
@ -892,11 +921,14 @@ void StorageFile::truncate(
    }
    else
    {
-        if (!fs::exists(paths[0]))
-            return;
+        for (const auto & path : paths)
+        {
+            if (!fs::exists(path))
+                continue;

-        if (0 != ::truncate(paths[0].c_str(), 0))
-            throwFromErrnoWithPath("Cannot truncate file " + paths[0], paths[0], ErrorCodes::CANNOT_TRUNCATE_FILE);
+            if (0 != ::truncate(path.c_str(), 0))
+                throwFromErrnoWithPath("Cannot truncate file " + path, path, ErrorCodes::CANNOT_TRUNCATE_FILE);
+        }
    }
 }

--- a/src/Storages/StorageFile.h
+++ b/src/Storages/StorageFile.h
@ -120,6 +120,8 @@ private:
    size_t total_bytes_to_read = 0;

    String path_for_partitioned_write;
+
+    bool is_path_with_globs = false;
 };

 }
--- a/src/Storages/StorageS3.cpp
+++ b/src/Storages/StorageS3.cpp
@ -68,7 +68,7 @@ namespace ErrorCodes
    extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
    extern const int S3_ERROR;
    extern const int UNEXPECTED_EXPRESSION;
-    extern const int CANNOT_OPEN_FILE;
+    extern const int DATABASE_ACCESS_DENIED;
    extern const int CANNOT_EXTRACT_TABLE_STRUCTURE;
 }

@ -82,8 +82,6 @@ public:
    Impl(Aws::S3::S3Client & client_, const S3::URI & globbed_uri_)
        : client(client_), globbed_uri(globbed_uri_)
    {
-        std::lock_guard lock(mutex);
-
        if (globbed_uri.bucket.find_first_of("*?{") != globbed_uri.bucket.npos)
            throw Exception("Expression can not have wildcards inside bucket name", ErrorCodes::UNEXPECTED_EXPRESSION);

@ -176,6 +174,37 @@ String StorageS3Source::DisclosedGlobIterator::next()
    return pimpl->next();
 }

+class StorageS3Source::KeysIterator::Impl
+{
+public:
+    explicit Impl(const std::vector<String> & keys_) : keys(keys_), keys_iter(keys.begin())
+    {
+    }
+
+    String next()
+    {
+        std::lock_guard lock(mutex);
+        if (keys_iter == keys.end())
+            return "";
+        auto key = *keys_iter;
+        ++keys_iter;
+        return key;
+    }
+
+private:
+    std::mutex mutex;
+    Strings keys;
+    Strings::iterator keys_iter;
+};
+
+StorageS3Source::KeysIterator::KeysIterator(const std::vector<String> & keys_) : pimpl(std::make_shared<StorageS3Source::KeysIterator::Impl>(keys_))
+{
+}
+
+String StorageS3Source::KeysIterator::next()
+{
+    return pimpl->next();
+}

 Block StorageS3Source::getHeader(Block sample_block, bool with_path_column, bool with_file_column)
 {
@ -296,6 +325,39 @@ Chunk StorageS3Source::generate()
    return generate();
 }

+static bool checkIfObjectExists(const std::shared_ptr<Aws::S3::S3Client> & client, const String & bucket, const String & key)
+{
+    bool is_finished = false;
+    Aws::S3::Model::ListObjectsV2Request request;
+    Aws::S3::Model::ListObjectsV2Outcome outcome;
+
+    request.SetBucket(bucket);
+    request.SetPrefix(key);
+    while (!is_finished)
+    {
+        outcome = client->ListObjectsV2(request);
+        if (!outcome.IsSuccess())
+            throw Exception(
+                ErrorCodes::S3_ERROR,
+                "Could not list objects in bucket {} with key {}, S3 exception: {}, message: {}",
+                quoteString(bucket),
+                quoteString(key),
+                backQuote(outcome.GetError().GetExceptionName()),
+                quoteString(outcome.GetError().GetMessage()));
+
+        const auto & result_batch = outcome.GetResult().GetContents();
+        for (const auto & obj : result_batch)
+        {
+            if (obj.GetKey() == key)
+                return true;
+        }
+
+        request.SetContinuationToken(outcome.GetResult().GetNextContinuationToken());
+        is_finished = !outcome.GetResult().GetIsTruncated();
+    }
+
+    return false;
+}

 class StorageS3Sink : public SinkToStorage
 {
@ -315,9 +377,6 @@ public:
        , sample_block(sample_block_)
        , format_settings(format_settings_)
    {
-        if (key.find_first_of("*?{") != std::string::npos)
-            throw Exception(ErrorCodes::CANNOT_OPEN_FILE, "S3 key '{}' contains globs, so the table is in readonly mode", key);
-
        write_buf = wrapWriteBufferWithCompressionMethod(
            std::make_unique<WriteBufferFromS3>(client, bucket, key, min_upload_part_size, max_single_part_upload_size), compression_method, 3);
        writer = FormatFactory::instance().getOutputFormatParallelIfPossible(format, *write_buf, sample_block, context, {}, format_settings);
@ -419,7 +478,6 @@ private:
    std::optional<FormatSettings> format_settings;

    ExpressionActionsPtr partition_by_expr;
-    String partition_by_column_name;

    static void validateBucket(const String & str)
    {
@ -468,6 +526,7 @@ StorageS3::StorageS3(
    ASTPtr partition_by_)
    : IStorage(table_id_)
    , client_auth{uri_, access_key_id_, secret_access_key_, max_connections_, {}, {}} /// Client and settings will be updated later
+    , keys({uri_.key})
    , format_name(format_name_)
    , max_single_read_retries(max_single_read_retries_)
    , min_upload_part_size(min_upload_part_size_)
@ -477,6 +536,7 @@ StorageS3::StorageS3(
    , distributed_processing(distributed_processing_)
    , format_settings(format_settings_)
    , partition_by(partition_by_)
+    , is_key_with_globs(uri_.key.find_first_of("*?{") != std::string::npos)
 {
    context_->getGlobalContext()->getRemoteHostFilter().checkURL(uri_.uri);
    StorageInMemoryMetadata storage_metadata;
@ -484,7 +544,7 @@ StorageS3::StorageS3(
    updateClientAndAuthSettings(context_, client_auth);
    if (columns_.empty())
    {
-        auto columns = getTableStructureFromDataImpl(format_name, client_auth, max_single_read_retries_, compression_method, distributed_processing_, format_settings, context_);
+        auto columns = getTableStructureFromDataImpl(format_name, client_auth, max_single_read_retries_, compression_method, distributed_processing_, is_key_with_globs, format_settings, context_);
        storage_metadata.setColumns(columns);
    }
    else
@ -495,9 +555,8 @@ StorageS3::StorageS3(
    setInMemoryMetadata(storage_metadata);
 }

-std::shared_ptr<StorageS3Source::IteratorWrapper> StorageS3::createFileIterator(const ClientAuthentication & client_auth, bool distributed_processing, ContextPtr local_context)
+std::shared_ptr<StorageS3Source::IteratorWrapper> StorageS3::createFileIterator(const ClientAuthentication & client_auth, const std::vector<String> & keys, bool is_key_with_globs, bool distributed_processing, ContextPtr local_context)
 {
-    std::shared_ptr<StorageS3Source::IteratorWrapper> iterator_wrapper{nullptr};
    if (distributed_processing)
    {
        return std::make_shared<StorageS3Source::IteratorWrapper>(
@ -505,13 +564,23 @@ std::shared_ptr<StorageS3Source::IteratorWrapper> StorageS3::createFileIterator(
                return callback();
        });
    }
-
-    /// Iterate through disclosed globs and make a source for each file
-    auto glob_iterator = std::make_shared<StorageS3Source::DisclosedGlobIterator>(*client_auth.client, client_auth.uri);
-    return std::make_shared<StorageS3Source::IteratorWrapper>([glob_iterator]()
+    else if (is_key_with_globs)
    {
-        return glob_iterator->next();
-    });
+        /// Iterate through disclosed globs and make a source for each file
+        auto glob_iterator = std::make_shared<StorageS3Source::DisclosedGlobIterator>(*client_auth.client, client_auth.uri);
+        return std::make_shared<StorageS3Source::IteratorWrapper>([glob_iterator]()
+        {
+            return glob_iterator->next();
+        });
+    }
+    else
+    {
+        auto keys_iterator = std::make_shared<StorageS3Source::KeysIterator>(keys);
+        return std::make_shared<StorageS3Source::IteratorWrapper>([keys_iterator]()
+        {
+            return keys_iterator->next();
+        });
+    }
 }

 Pipe StorageS3::read(
@ -536,7 +605,7 @@ Pipe StorageS3::read(
            need_file_column = true;
    }

-    std::shared_ptr<StorageS3Source::IteratorWrapper> iterator_wrapper = createFileIterator(client_auth, distributed_processing, local_context);
+    std::shared_ptr<StorageS3Source::IteratorWrapper> iterator_wrapper = createFileIterator(client_auth, keys, is_key_with_globs, distributed_processing, local_context);

    for (size_t i = 0; i < num_streams; ++i)
    {
@ -567,8 +636,8 @@ SinkToStoragePtr StorageS3::write(const ASTPtr & query, const StorageMetadataPtr
    updateClientAndAuthSettings(local_context, client_auth);

    auto sample_block = metadata_snapshot->getSampleBlock();
-    auto chosen_compression_method = chooseCompressionMethod(client_auth.uri.key, compression_method);
-    bool has_wildcards = client_auth.uri.bucket.find(PARTITION_ID_WILDCARD) != String::npos || client_auth.uri.key.find(PARTITION_ID_WILDCARD) != String::npos;
+    auto chosen_compression_method = chooseCompressionMethod(keys.back(), compression_method);
+    bool has_wildcards = client_auth.uri.bucket.find(PARTITION_ID_WILDCARD) != String::npos || keys.back().find(PARTITION_ID_WILDCARD) != String::npos;
    auto insert_query = std::dynamic_pointer_cast<ASTInsertQuery>(query);

    auto partition_by_ast = insert_query ? (insert_query->partition_by ? insert_query->partition_by : partition_by) : nullptr;
@ -585,12 +654,41 @@ SinkToStoragePtr StorageS3::write(const ASTPtr & query, const StorageMetadataPtr
            chosen_compression_method,
            client_auth.client,
            client_auth.uri.bucket,
-            client_auth.uri.key,
+            keys.back(),
            min_upload_part_size,
            max_single_part_upload_size);
    }
    else
    {
+        if (is_key_with_globs)
+            throw Exception(ErrorCodes::DATABASE_ACCESS_DENIED, "S3 key '{}' contains globs, so the table is in readonly mode", client_auth.uri.key);
+
+        bool truncate_in_insert = local_context->getSettingsRef().s3_truncate_on_insert;
+
+        if (!truncate_in_insert && checkIfObjectExists(client_auth.client, client_auth.uri.bucket, keys.back()))
+        {
+            if (local_context->getSettingsRef().s3_create_new_file_on_insert)
+            {
+                size_t index = keys.size();
+                auto pos = keys[0].find_first_of('.');
+                String new_key;
+                do
+                {
+                    new_key = keys[0].substr(0, pos) + "." + std::to_string(index) + (pos == std::string::npos ? "" : keys[0].substr(pos));
+                    ++index;
+                }
+                while (checkIfObjectExists(client_auth.client, client_auth.uri.bucket, new_key));
+                keys.push_back(new_key);
+            }
+            else
+                throw Exception(
+                    ErrorCodes::BAD_ARGUMENTS,
+                    "Object in bucket {} with key {} already exists. If you want to overwrite it, enable setting s3_truncate_on_insert, if you "
+                    "want to create a new file on each insert, enable setting s3_create_new_file_on_insert",
+                    client_auth.uri.bucket,
+                    keys.back());
+        }
+
        return std::make_shared<StorageS3Sink>(
            format_name,
            sample_block,
@ -599,7 +697,7 @@ SinkToStoragePtr StorageS3::write(const ASTPtr & query, const StorageMetadataPtr
            chosen_compression_method,
            client_auth.client,
            client_auth.uri.bucket,
-            client_auth.uri.key,
+            keys.back(),
            min_upload_part_size,
            max_single_part_upload_size);
    }
@ -610,11 +708,17 @@ void StorageS3::truncate(const ASTPtr & /* query */, const StorageMetadataPtr &,
 {
    updateClientAndAuthSettings(local_context, client_auth);

-    Aws::S3::Model::ObjectIdentifier obj;
-    obj.SetKey(client_auth.uri.key);
+    if (is_key_with_globs)
+        throw Exception(ErrorCodes::DATABASE_ACCESS_DENIED, "S3 key '{}' contains globs, so the table is in readonly mode", client_auth.uri.key);

    Aws::S3::Model::Delete delkeys;
-    delkeys.AddObjects(std::move(obj));
+
+    for (const auto & key : keys)
+    {
+        Aws::S3::Model::ObjectIdentifier obj;
+        obj.SetKey(key);
+        delkeys.AddObjects(std::move(obj));
+    }

    Aws::S3::Model::DeleteObjectsRequest request;
    request.SetBucket(client_auth.uri.bucket);
@ -734,7 +838,7 @@ ColumnsDescription StorageS3::getTableStructureFromData(
 {
    ClientAuthentication client_auth{uri, access_key_id, secret_access_key, max_connections, {}, {}};
    updateClientAndAuthSettings(ctx, client_auth);
-    return getTableStructureFromDataImpl(format, client_auth, max_single_read_retries, compression_method, distributed_processing, format_settings, ctx);
+    return getTableStructureFromDataImpl(format, client_auth, max_single_read_retries, compression_method, distributed_processing, uri.key.find_first_of("*?{") != std::string::npos, format_settings, ctx);
 }

 ColumnsDescription StorageS3::getTableStructureFromDataImpl(
@ -743,12 +847,14 @@ ColumnsDescription StorageS3::getTableStructureFromDataImpl(
    UInt64 max_single_read_retries,
    const String & compression_method,
    bool distributed_processing,
+    bool is_key_with_globs,
    const std::optional<FormatSettings> & format_settings,
    ContextPtr ctx)
 {
+    std::vector<String> keys = {client_auth.uri.key};
    auto read_buffer_creator = [&]()
    {
-        auto file_iterator = createFileIterator(client_auth, distributed_processing, ctx);
+        auto file_iterator = createFileIterator(client_auth, keys, is_key_with_globs, distributed_processing, ctx);
        String current_key = (*file_iterator)();
        if (current_key.empty())
            throw Exception(
--- a/src/Storages/StorageS3.h
+++ b/src/Storages/StorageS3.h
@ -44,6 +44,18 @@ public:
            std::shared_ptr<Impl> pimpl;
    };

+    class KeysIterator
+    {
+        public:
+            explicit KeysIterator(const std::vector<String> & keys_);
+            String next();
+
+        private:
+            class Impl;
+            /// shared_ptr to have copy constructor
+            std::shared_ptr<Impl> pimpl;
+    };
+
    using IteratorWrapper = std::function<String()>;

    static Block getHeader(Block sample_block, bool with_path_column, bool with_file_column);
@ -174,6 +186,7 @@ private:
    };

    ClientAuthentication client_auth;
+    std::vector<String> keys;

    String format_name;
    UInt64 max_single_read_retries;
@ -184,10 +197,11 @@ private:
    const bool distributed_processing;
    std::optional<FormatSettings> format_settings;
    ASTPtr partition_by;
+    bool is_key_with_globs = false;

    static void updateClientAndAuthSettings(ContextPtr, ClientAuthentication &);

-    static std::shared_ptr<StorageS3Source::IteratorWrapper> createFileIterator(const ClientAuthentication & client_auth, bool distributed_processing, ContextPtr local_context);
+    static std::shared_ptr<StorageS3Source::IteratorWrapper> createFileIterator(const ClientAuthentication & client_auth, const std::vector<String> & keys, bool is_key_with_globs, bool distributed_processing, ContextPtr local_context);

    static ColumnsDescription getTableStructureFromDataImpl(
        const String & format,
@ -195,6 +209,7 @@ private:
        UInt64 max_single_read_retries,
        const String & compression_method,
        bool distributed_processing,
+        bool is_key_with_globs,
        const std::optional<FormatSettings> & format_settings,
        ContextPtr ctx);
 };
--- a/tests/ci/docker_images_check.py
+++ b/tests/ci/docker_images_check.py
@ -6,7 +6,7 @@ import os
 import shutil
 import subprocess
 import time
-from typing import List, Tuple
+from typing import List, Optional, Set, Tuple, Union

 from github import Github

@ -24,13 +24,54 @@ NAME = "Push to Dockerhub (actions)"
 TEMP_PATH = os.path.join(RUNNER_TEMP, "docker_images_check")


+class DockerImage:
+    def __init__(
+        self,
+        path: str,
+        repo: str,
+        parent: Optional["DockerImage"] = None,
+        gh_repo_path: str = GITHUB_WORKSPACE,
+    ):
+        self.path = path
+        self.full_path = os.path.join(gh_repo_path, path)
+        self.repo = repo
+        self.parent = parent
+        self.built = False
+
+    def __eq__(self, other) -> bool:  # type: ignore
+        """Is used to check if DockerImage is in a set or not"""
+        return self.path == other.path and self.repo == self.repo
+
+    def __lt__(self, other) -> bool:
+        if not isinstance(other, DockerImage):
+            return False
+        if self.parent and not other.parent:
+            return False
+        if not self.parent and other.parent:
+            return True
+        if self.path < other.path:
+            return True
+        if self.repo < other.repo:
+            return True
+        return False
+
+    def __hash__(self):
+        return hash(self.path)
+
+    def __str__(self):
+        return self.repo
+
+    def __repr__(self):
+        return f"DockerImage(path={self.path},repo={self.repo},parent={self.parent})"
+
+
 def get_changed_docker_images(
    pr_info: PRInfo, repo_path: str, image_file_path: str
-) -> List[Tuple[str, str]]:
+) -> Set[DockerImage]:
    images_dict = {}
    path_to_images_file = os.path.join(repo_path, image_file_path)
    if os.path.exists(path_to_images_file):
-        with open(path_to_images_file, "r") as dict_file:
+        with open(path_to_images_file, "rb") as dict_file:
            images_dict = json.load(dict_file)
    else:
        logging.info(
@ -38,7 +79,7 @@ def get_changed_docker_images(
        )

    if not images_dict:
-        return []
+        return set()

    files_changed = pr_info.changed_files

@ -54,14 +95,15 @@ def get_changed_docker_images(
    for dockerfile_dir, image_description in images_dict.items():
        for f in files_changed:
            if f.startswith(dockerfile_dir):
+                name = image_description["name"]
                logging.info(
                    "Found changed file '%s' which affects "
                    "docker image '%s' with path '%s'",
                    f,
-                    image_description["name"],
+                    name,
                    dockerfile_dir,
                )
-                changed_images.append(dockerfile_dir)
+                changed_images.append(DockerImage(dockerfile_dir, name))
                break

    # The order is important: dependents should go later than bases, so that
@ -69,14 +111,16 @@ def get_changed_docker_images(
    index = 0
    while index < len(changed_images):
        image = changed_images[index]
-        for dependent in images_dict[image]["dependent"]:
+        for dependent in images_dict[image.path]["dependent"]:
            logging.info(
                "Marking docker image '%s' as changed because it "
                "depends on changed docker image '%s'",
                dependent,
                image,
            )
-            changed_images.append(dependent)
+            changed_images.append(
+                DockerImage(dependent, images_dict[dependent]["name"], image)
+            )
        index += 1
        if index > 5 * len(images_dict):
            # Sanity check to prevent infinite loop.
@ -84,19 +128,9 @@ def get_changed_docker_images(
                f"Too many changed docker images, this is a bug. {changed_images}"
            )

-    # If a dependent image was already in the list because its own files
-    # changed, but then it was added as a dependent of a changed base, we
-    # must remove the earlier entry so that it doesn't go earlier than its
-    # base. This way, the dependent will be rebuilt later than the base, and
-    # will correctly use the updated version of the base.
-    seen = set()
-    no_dups_reversed = []
-    for x in reversed(changed_images):
-        if x not in seen:
-            seen.add(x)
-            no_dups_reversed.append(x)
-
-    result = [(x, images_dict[x]["name"]) for x in reversed(no_dups_reversed)]
+    # With reversed changed_images set will use images with parents first, and
+    # images without parents then
+    result = set(reversed(changed_images))
    logging.info(
        "Changed docker images for PR %s @ %s: '%s'",
        pr_info.number,
@ -106,66 +140,109 @@ def get_changed_docker_images(
    return result


+def gen_versions(
+    pr_info: PRInfo, suffix: Optional[str]
+) -> Tuple[List[str], Union[str, List[str]]]:
+    pr_commit_version = str(pr_info.number) + "-" + pr_info.sha
+    # The order is important, PR number is used as cache during the build
+    versions = [str(pr_info.number), pr_commit_version]
+    result_version = pr_commit_version
+    if pr_info.number == 0:
+        # First get the latest for cache
+        versions.insert(0, "latest")
+
+    if suffix:
+        # We should build architecture specific images separately and merge a
+        # manifest lately in a different script
+        versions = [f"{v}-{suffix}" for v in versions]
+        # changed_images_{suffix}.json should contain all changed images
+        result_version = versions
+
+    return versions, result_version
+
+
 def build_and_push_one_image(
-    path_to_dockerfile_folder: str, image_name: str, version_string: str, push: bool
+    image: DockerImage,
+    version_string: str,
+    push: bool,
+    child: bool,
 ) -> Tuple[bool, str]:
-    path = path_to_dockerfile_folder
    logging.info(
        "Building docker image %s with version %s from path %s",
-        image_name,
+        image.repo,
        version_string,
-        path,
+        image.full_path,
    )
    build_log = os.path.join(
-        TEMP_PATH,
-        "build_and_push_log_{}_{}".format(
-            str(image_name).replace("/", "_"), version_string
-        ),
+        TEMP_PATH, f"build_and_push_log_{image.repo.replace('/', '_')}_{version_string}"
    )
    push_arg = ""
    if push:
        push_arg = "--push "

-    with open(build_log, "w") as bl:
+    from_tag_arg = ""
+    if child:
+        from_tag_arg = f"--build-arg FROM_TAG={version_string} "
+
+    with open(build_log, "wb") as bl:
        cmd = (
            "docker buildx build --builder default "
-            f"--build-arg FROM_TAG={version_string} "
+            f"{from_tag_arg}"
            f"--build-arg BUILDKIT_INLINE_CACHE=1 "
-            f"--tag {image_name}:{version_string} "
-            f"--cache-from type=registry,ref={image_name}:{version_string} "
+            f"--tag {image.repo}:{version_string} "
+            f"--cache-from type=registry,ref={image.repo}:{version_string} "
            f"{push_arg}"
-            f"--progress plain {path}"
+            f"--progress plain {image.full_path}"
        )
        logging.info("Docker command to run: %s", cmd)
-        retcode = subprocess.Popen(cmd, shell=True, stderr=bl, stdout=bl).wait()
+        with subprocess.Popen(cmd, shell=True, stderr=bl, stdout=bl) as proc:
+            retcode = proc.wait()
+
        if retcode != 0:
            return False, build_log

-    logging.info("Processing of %s successfully finished", image_name)
+    logging.info("Processing of %s successfully finished", image.repo)
    return True, build_log


 def process_single_image(
-    versions: List[str], path_to_dockerfile_folder: str, image_name: str, push: bool
+    image: DockerImage,
+    versions: List[str],
+    push: bool,
+    child: bool,
 ) -> List[Tuple[str, str, str]]:
    logging.info("Image will be pushed with versions %s", ", ".join(versions))
    result = []
    for ver in versions:
        for i in range(5):
-            success, build_log = build_and_push_one_image(
-                path_to_dockerfile_folder, image_name, ver, push
-            )
+            success, build_log = build_and_push_one_image(image, ver, push, child)
            if success:
-                result.append((image_name + ":" + ver, build_log, "OK"))
+                result.append((image.repo + ":" + ver, build_log, "OK"))
                break
            logging.info(
                "Got error will retry %s time and sleep for %s seconds", i, i * 5
            )
            time.sleep(i * 5)
        else:
-            result.append((image_name + ":" + ver, build_log, "FAIL"))
+            result.append((image.repo + ":" + ver, build_log, "FAIL"))

    logging.info("Processing finished")
+    image.built = True
+    return result
+
+
+def process_image_with_parents(
+    image: DockerImage, versions: List[str], push: bool, child: bool = False
+) -> List[Tuple[str, str, str]]:
+    result = []  # type: List[Tuple[str,str,str]]
+    if image.built:
+        return result
+
+    if image.parent is not None:
+        result += process_image_with_parents(image.parent, versions, push, False)
+        child = True
+
+    result += process_single_image(image, versions, push, child)
    return result


@ -182,7 +259,7 @@ def process_test_results(
            build_url = s3_client.upload_test_report_to_s3(
                build_log, s3_path_prefix + "/" + os.path.basename(build_log)
            )
-            url_part += '<a href="{}">build_log</a>'.format(build_url)
+            url_part += f'<a href="{build_url}">build_log</a>'
        if url_part:
            test_name = image + " (" + url_part + ")"
        else:
@ -255,8 +332,6 @@ def main():
            shell=True,
        )

-    repo_path = GITHUB_WORKSPACE
-
    if os.path.exists(TEMP_PATH):
        shutil.rmtree(TEMP_PATH)
    os.makedirs(TEMP_PATH)
@ -267,43 +342,30 @@ def main():
    else:
        pr_info = PRInfo(need_changed_files=True)

-    changed_images = get_changed_docker_images(pr_info, repo_path, "docker/images.json")
-    logging.info(
-        "Has changed images %s", ", ".join([str(image[0]) for image in changed_images])
+    changed_images = get_changed_docker_images(
+        pr_info, GITHUB_WORKSPACE, "docker/images.json"
    )
-    pr_commit_version = str(pr_info.number) + "-" + pr_info.sha
-    # The order is important, PR number is used as cache during the build
-    versions = [str(pr_info.number), pr_commit_version]
-    result_version = pr_commit_version
-    if pr_info.number == 0:
-        # First get the latest for cache
-        versions.insert(0, "latest")
+    logging.info("Has changed images %s", ", ".join([im.path for im in changed_images]))

-    if args.suffix:
-        # We should build architecture specific images separately and merge a
-        # manifest lately in a different script
-        versions = [f"{v}-{args.suffix}" for v in versions]
-        # changed_images_{suffix}.json should contain all changed images
-        result_version = versions
+    image_versions, result_version = gen_versions(pr_info, args.suffix)

    result_images = {}
    images_processing_result = []
-    for rel_path, image_name in changed_images:
-        full_path = os.path.join(repo_path, rel_path)
-        images_processing_result += process_single_image(
-            versions, full_path, image_name, push
+    for image in changed_images:
+        images_processing_result += process_image_with_parents(
+            image, image_versions, push
        )
-        result_images[image_name] = result_version
+        result_images[image.repo] = result_version

    if changed_images:
-        description = "Updated " + ",".join([im[1] for im in changed_images])
+        description = "Updated " + ",".join([im.repo for im in changed_images])
    else:
        description = "Nothing to update"

    if len(description) >= 140:
        description = description[:136] + "..."

-    with open(changed_json, "w") as images_file:
+    with open(changed_json, "w", encoding="utf-8") as images_file:
        json.dump(result_images, images_file)

    s3_helper = S3Helper("https://s3.amazonaws.com")
@ -317,8 +379,8 @@ def main():

    url = upload_results(s3_helper, pr_info.number, pr_info.sha, test_results, [], NAME)

-    print("::notice ::Report url: {}".format(url))
-    print('::set-output name=url_output::"{}"'.format(url))
+    print(f"::notice ::Report url: {url}")
+    print(f'::set-output name=url_output::"{url}"')

    if args.no_reports:
        return
--- a/tests/ci/docker_test.py
+++ b/tests/ci/docker_test.py
@ -0,0 +1,171 @@
+#!/usr/bin/env python
+
+import os
+import unittest
+from unittest.mock import patch
+
+from pr_info import PRInfo
+import docker_images_check as di
+
+# di.logging.basicConfig(level=di.logging.INFO)
+
+
+class TestDockerImageCheck(unittest.TestCase):
+    docker_images_path = os.path.join(
+        os.path.dirname(__file__), "tests/docker_images.json"
+    )
+
+    def test_get_changed_docker_images(self):
+        pr_info = PRInfo(PRInfo.default_event.copy())
+        pr_info.changed_files = {
+            "docker/test/stateless",
+            "docker/test/base",
+            "docker/docs/builder",
+        }
+        images = sorted(
+            list(di.get_changed_docker_images(pr_info, "/", self.docker_images_path))
+        )
+        self.maxDiff = None
+        expected = sorted(
+            [
+                di.DockerImage("docker/test/base", "clickhouse/test-base"),
+                di.DockerImage("docker/docs/builder", "clickhouse/docs-builder"),
+                di.DockerImage(
+                    "docker/test/stateless",
+                    "clickhouse/stateless-test",
+                    "clickhouse/test-base",
+                ),
+                di.DockerImage(
+                    "docker/test/integration/base",
+                    "clickhouse/integration-test",
+                    "clickhouse/test-base",
+                ),
+                di.DockerImage(
+                    "docker/test/fuzzer", "clickhouse/fuzzer", "clickhouse/test-base"
+                ),
+                di.DockerImage(
+                    "docker/test/keeper-jepsen",
+                    "clickhouse/keeper-jepsen-test",
+                    "clickhouse/test-base",
+                ),
+                di.DockerImage(
+                    "docker/docs/check",
+                    "clickhouse/docs-check",
+                    "clickhouse/docs-builder",
+                ),
+                di.DockerImage(
+                    "docker/docs/release",
+                    "clickhouse/docs-release",
+                    "clickhouse/docs-builder",
+                ),
+                di.DockerImage(
+                    "docker/test/stateful",
+                    "clickhouse/stateful-test",
+                    "clickhouse/stateless-test",
+                ),
+                di.DockerImage(
+                    "docker/test/unit",
+                    "clickhouse/unit-test",
+                    "clickhouse/stateless-test",
+                ),
+                di.DockerImage(
+                    "docker/test/stress",
+                    "clickhouse/stress-test",
+                    "clickhouse/stateful-test",
+                ),
+            ]
+        )
+        self.assertEqual(images, expected)
+
+    def test_gen_version(self):
+        pr_info = PRInfo(PRInfo.default_event.copy())
+        versions, result_version = di.gen_versions(pr_info, None)
+        self.assertEqual(versions, ["latest", "0", "0-HEAD"])
+        self.assertEqual(result_version, "0-HEAD")
+        versions, result_version = di.gen_versions(pr_info, "suffix")
+        self.assertEqual(versions, ["latest-suffix", "0-suffix", "0-HEAD-suffix"])
+        self.assertEqual(result_version, versions)
+        pr_info.number = 1
+        versions, result_version = di.gen_versions(pr_info, None)
+        self.assertEqual(versions, ["1", "1-HEAD"])
+        self.assertEqual(result_version, "1-HEAD")
+
+    @patch("builtins.open")
+    @patch("subprocess.Popen")
+    def test_build_and_push_one_image(self, mock_popen, mock_open):
+        mock_popen.return_value.__enter__.return_value.wait.return_value = 0
+        image = di.DockerImage("path", "name", gh_repo_path="")
+
+        result, _ = di.build_and_push_one_image(image, "version", True, True)
+        mock_open.assert_called_once()
+        mock_popen.assert_called_once()
+        self.assertIn(
+            "docker buildx build --builder default --build-arg FROM_TAG=version "
+            "--build-arg BUILDKIT_INLINE_CACHE=1 --tag name:version --cache-from "
+            "type=registry,ref=name:version --push --progress plain path",
+            mock_popen.call_args.args,
+        )
+        self.assertTrue(result)
+
+        mock_open.reset()
+        mock_popen.reset()
+        mock_popen.return_value.__enter__.return_value.wait.return_value = 0
+        result, _ = di.build_and_push_one_image(image, "version2", False, True)
+        self.assertIn(
+            "docker buildx build --builder default --build-arg FROM_TAG=version2 "
+            "--build-arg BUILDKIT_INLINE_CACHE=1 --tag name:version2 --cache-from "
+            "type=registry,ref=name:version2 --progress plain path",
+            mock_popen.call_args.args,
+        )
+        self.assertTrue(result)
+
+        mock_popen.return_value.__enter__.return_value.wait.return_value = 1
+        result, _ = di.build_and_push_one_image(image, "version2", False, False)
+        self.assertIn(
+            "docker buildx build --builder default "
+            "--build-arg BUILDKIT_INLINE_CACHE=1 --tag name:version2 --cache-from "
+            "type=registry,ref=name:version2 --progress plain path",
+            mock_popen.call_args.args,
+        )
+        self.assertFalse(result)
+
+    @patch("docker_images_check.build_and_push_one_image")
+    def test_process_image_with_parents(self, mock_build):
+        mock_build.side_effect = lambda w, x, y, z: (True, f"{w.repo}_{x}.log")
+        im1 = di.DockerImage("path1", "repo1")
+        im2 = di.DockerImage("path2", "repo2", im1)
+        im3 = di.DockerImage("path3", "repo3", im2)
+        im4 = di.DockerImage("path4", "repo4", im1)
+        # We use list to have determined order of image builgings
+        images = [im4, im1, im3, im2, im1]
+        results = [
+            di.process_image_with_parents(im, ["v1", "v2", "latest"], True)
+            for im in images
+        ]
+
+        expected = [
+            [  # repo4 -> repo1
+                ("repo1:v1", "repo1_v1.log", "OK"),
+                ("repo1:v2", "repo1_v2.log", "OK"),
+                ("repo1:latest", "repo1_latest.log", "OK"),
+                ("repo4:v1", "repo4_v1.log", "OK"),
+                ("repo4:v2", "repo4_v2.log", "OK"),
+                ("repo4:latest", "repo4_latest.log", "OK"),
+            ],
+            [],  # repo1 is built
+            [  # repo3 -> repo2 -> repo1
+                ("repo2:v1", "repo2_v1.log", "OK"),
+                ("repo2:v2", "repo2_v2.log", "OK"),
+                ("repo2:latest", "repo2_latest.log", "OK"),
+                ("repo3:v1", "repo3_v1.log", "OK"),
+                ("repo3:v2", "repo3_v2.log", "OK"),
+                ("repo3:latest", "repo3_latest.log", "OK"),
+            ],
+            [],  # repo2 -> repo1 are built
+            [],  # repo1 is built
+        ]
+        self.assertEqual(results, expected)
+
+
+if __name__ == "__main__":
+    unittest.main()
--- a/tests/ci/metrics_lambda/app.py
+++ b/tests/ci/metrics_lambda/app.py
@ -174,6 +174,7 @@ def group_runners_by_tag(listed_runners):
        "fuzzer-unit-tester",
        "stress-tester",
        "style-checker",
+        "style-checker-aarch64",
    ]
    for runner in listed_runners:
        for tag in runner.tags:
--- a/tests/ci/pr_info.py
+++ b/tests/ci/pr_info.py
@ -11,6 +11,8 @@ DIFF_IN_DOCUMENTATION_EXT = [".html", ".md", ".yml", ".txt", ".css", ".js", ".xm
                             ".jpg", ".py", ".sh", ".json"]

 def get_pr_for_commit(sha, ref):
+    if not ref:
+        return None
    try_get_pr_url = f"https://api.github.com/repos/{GITHUB_REPOSITORY}/commits/{sha}/pulls"
    try:
        response = requests.get(try_get_pr_url)
@ -32,23 +34,24 @@ def get_pr_for_commit(sha, ref):


 class PRInfo:
+    default_event = {
+                    'commits': 1,
+                    'before': 'HEAD~',
+                    'after': 'HEAD',
+                    'ref': None,
+                }
    def __init__(self, github_event=None, need_orgs=False, need_changed_files=False, labels_from_api=False):
        if not github_event:
            if GITHUB_EVENT_PATH:
                with open(GITHUB_EVENT_PATH, 'r', encoding='utf-8') as event_file:
                    github_event = json.load(event_file)
            else:
-                github_event = {
-                    'commits': 1,
-                    'before': 'HEAD~',
-                    'after': 'HEAD',
-                    'ref': None,
-                }
+                github_event = PRInfo.default_event.copy()
        self.event = github_event
        self.changed_files = set([])
        self.body = ""
        ref = github_event.get("ref", "refs/head/master")
-        if ref.startswith('refs/heads/'):
+        if ref and ref.startswith('refs/heads/'):
            ref = ref[11:]

        # workflow completed event, used for PRs only
--- a/tests/ci/tests/docker_images.json
+++ b/tests/ci/tests/docker_images.json
@ -0,0 +1,166 @@
+{
+    "docker/packager/deb": {
+        "name": "clickhouse/deb-builder",
+        "dependent": []
+    },
+    "docker/packager/binary": {
+        "name": "clickhouse/binary-builder",
+        "dependent": [
+            "docker/test/split_build_smoke_test",
+            "docker/test/pvs",
+            "docker/test/codebrowser"
+        ]
+    },
+    "docker/test/compatibility/centos": {
+        "name": "clickhouse/test-old-centos",
+        "dependent": []
+    },
+    "docker/test/compatibility/ubuntu": {
+        "name": "clickhouse/test-old-ubuntu",
+        "dependent": []
+    },
+    "docker/test/integration/base": {
+        "name": "clickhouse/integration-test",
+        "dependent": []
+    },
+    "docker/test/fuzzer": {
+        "name": "clickhouse/fuzzer",
+        "dependent": []
+    },
+    "docker/test/performance-comparison": {
+        "name": "clickhouse/performance-comparison",
+        "dependent": []
+    },
+    "docker/test/pvs": {
+        "name": "clickhouse/pvs-test",
+        "dependent": []
+    },
+    "docker/test/util": {
+        "name": "clickhouse/test-util",
+        "dependent": [
+            "docker/test/base",
+            "docker/test/fasttest"
+        ]
+    },
+    "docker/test/stateless": {
+        "name": "clickhouse/stateless-test",
+        "dependent": [
+            "docker/test/stateful",
+            "docker/test/unit"
+        ]
+    },
+    "docker/test/stateful": {
+        "name": "clickhouse/stateful-test",
+        "dependent": [
+            "docker/test/stress"
+        ]
+    },
+    "docker/test/unit": {
+        "name": "clickhouse/unit-test",
+        "dependent": []
+    },
+    "docker/test/stress": {
+        "name": "clickhouse/stress-test",
+        "dependent": []
+    },
+    "docker/test/split_build_smoke_test": {
+        "name": "clickhouse/split-build-smoke-test",
+        "dependent": []
+    },
+    "docker/test/codebrowser": {
+        "name": "clickhouse/codebrowser",
+        "dependent": []
+    },
+    "docker/test/integration/runner": {
+        "name": "clickhouse/integration-tests-runner",
+        "dependent": []
+    },
+    "docker/test/testflows/runner": {
+        "name": "clickhouse/testflows-runner",
+        "dependent": []
+    },
+    "docker/test/fasttest": {
+        "name": "clickhouse/fasttest",
+        "dependent": []
+    },
+    "docker/test/style": {
+        "name": "clickhouse/style-test",
+        "dependent": []
+    },
+    "docker/test/integration/s3_proxy": {
+        "name": "clickhouse/s3-proxy",
+        "dependent": []
+    },
+    "docker/test/integration/resolver": {
+        "name": "clickhouse/python-bottle",
+        "dependent": []
+    },
+    "docker/test/integration/helper_container": {
+        "name": "clickhouse/integration-helper",
+        "dependent": []
+    },
+    "docker/test/integration/mysql_golang_client": {
+        "name": "clickhouse/mysql-golang-client",
+        "dependent": []
+    },
+    "docker/test/integration/dotnet_client": {
+        "name": "clickhouse/dotnet-client",
+        "dependent": []
+    },
+    "docker/test/integration/mysql_java_client": {
+        "name": "clickhouse/mysql-java-client",
+        "dependent": []
+    },
+    "docker/test/integration/mysql_js_client": {
+        "name": "clickhouse/mysql-js-client",
+        "dependent": []
+    },
+    "docker/test/integration/mysql_php_client": {
+        "name": "clickhouse/mysql-php-client",
+        "dependent": []
+    },
+    "docker/test/integration/postgresql_java_client": {
+        "name": "clickhouse/postgresql-java-client",
+        "dependent": []
+    },
+    "docker/test/integration/kerberos_kdc": {
+        "name": "clickhouse/kerberos-kdc",
+        "dependent": []
+    },
+    "docker/test/base": {
+         "name": "clickhouse/test-base",
+         "dependent": [
+            "docker/test/stateless",
+            "docker/test/integration/base",
+            "docker/test/fuzzer",
+            "docker/test/keeper-jepsen"
+         ]
+    },
+    "docker/test/integration/kerberized_hadoop": {
+        "name": "clickhouse/kerberized-hadoop",
+        "dependent": []
+    },
+    "docker/test/sqlancer": {
+        "name": "clickhouse/sqlancer-test",
+        "dependent": []
+    },
+    "docker/test/keeper-jepsen": {
+        "name": "clickhouse/keeper-jepsen-test",
+        "dependent": []
+    },
+    "docker/docs/builder": {
+        "name": "clickhouse/docs-builder",
+        "dependent": [
+            "docker/docs/check",
+            "docker/docs/release"
+        ]
+    },
+    "docker/docs/check": {
+        "name": "clickhouse/docs-check",
+        "dependent": []
+    },
+    "docker/docs/release": {
+        "name": "clickhouse/docs-release",
+        "dependent": []
+    }
+}
--- a/tests/integration/CMakeLists.txt
+++ b/tests/integration/CMakeLists.txt
@ -10,7 +10,7 @@ find_program(PYTEST_CMD pytest)
 find_program(SUDO_CMD sudo)

 # will mount only one binary to docker container - build with .so cant work
-if(MAKE_STATIC_LIBRARIES AND DOCKER_CMD)
+if(USE_STATIC_LIBRARIES AND DOCKER_CMD)
    if(INTEGRATION_USE_RUNNER AND SUDO_CMD)
        add_test(NAME integration-runner WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR} COMMAND ${SUDO_CMD} ${CMAKE_CURRENT_SOURCE_DIR}/runner --binary ${ClickHouse_BINARY_DIR}/programs/clickhouse --configs-dir ${ClickHouse_SOURCE_DIR}/programs/server/)
        message(STATUS "Using tests in docker with runner SUDO=${SUDO_CMD}; DOCKER=${DOCKER_CMD};")
--- a/tests/integration/test_storage_hdfs/test.py
+++ b/tests/integration/test_storage_hdfs/test.py
@ -366,6 +366,43 @@ def test_hdfs_directory_not_exist(started_cluster):
    node1.query(ddl)
    assert "" == node1.query("select * from HDFSStorageWithNotExistDir")

+def test_overwrite(started_cluster):
+    hdfs_api = started_cluster.hdfs_api
+
+    table_function = f"hdfs('hdfs://hdfs1:9000/data', 'Parquet', 'a Int32, b String')"
+    node1.query(f"create table test_overwrite as {table_function}")
+    node1.query(f"insert into test_overwrite select number, randomString(100) from numbers(5)")
+    node1.query_and_get_error(f"insert into test_overwrite select number, randomString(100) FROM numbers(10)")
+    node1.query(f"insert into test_overwrite select number, randomString(100) from numbers(10) settings hdfs_truncate_on_insert=1")
+
+    result = node1.query(f"select count() from test_overwrite")
+    assert(int(result) == 10)
+
+
+def test_multiple_inserts(started_cluster):
+    hdfs_api = started_cluster.hdfs_api
+
+    table_function = f"hdfs('hdfs://hdfs1:9000/data_multiple_inserts', 'Parquet', 'a Int32, b String')"
+    node1.query(f"create table test_multiple_inserts as {table_function}")
+    node1.query(f"insert into test_multiple_inserts select number, randomString(100) from numbers(10)")
+    node1.query(f"insert into test_multiple_inserts select number, randomString(100) from numbers(20) settings hdfs_create_new_file_on_insert=1")
+    node1.query(f"insert into test_multiple_inserts select number, randomString(100) from numbers(30) settings hdfs_create_new_file_on_insert=1")
+
+    result = node1.query(f"select count() from test_multiple_inserts")
+    assert(int(result) == 60)
+
+    result = node1.query(f"drop table test_multiple_inserts")
+
+    table_function = f"hdfs('hdfs://hdfs1:9000/data_multiple_inserts.gz', 'Parquet', 'a Int32, b String')"
+    node1.query(f"create table test_multiple_inserts as {table_function}")
+    node1.query(f"insert into test_multiple_inserts select number, randomString(100) FROM numbers(10)")
+    node1.query(f"insert into test_multiple_inserts select number, randomString(100) FROM numbers(20) settings hdfs_create_new_file_on_insert=1")
+    node1.query(f"insert into test_multiple_inserts select number, randomString(100) FROM numbers(30) settings hdfs_create_new_file_on_insert=1")
+
+    result = node1.query(f"select count() from test_multiple_inserts")
+    assert(int(result) == 60)
+
+    
 def test_format_detection(started_cluster):
    node1.query(f"create table arrow_table (x UInt64) engine=HDFS('hdfs://hdfs1:9000/data.arrow')")
    node1.query(f"insert into arrow_table select 1")
--- a/tests/integration/test_storage_s3/configs/named_collections.xml
+++ b/tests/integration/test_storage_s3/configs/named_collections.xml
@ -10,6 +10,11 @@
            <access_key_id>minio</access_key_id>
            <secret_access_key>minio123</secret_access_key>
        </s3_parquet>
+        <s3_parquet_gz>
+            <url>http://minio1:9001/root/test_parquet_gz</url>
+            <access_key_id>minio</access_key_id>
+            <secret_access_key>minio123</secret_access_key>
+        </s3_parquet_gz>
        <s3_orc>
            <url>http://minio1:9001/root/test_orc</url>
            <access_key_id>minio</access_key_id>
--- a/tests/integration/test_storage_s3/test.py
+++ b/tests/integration/test_storage_s3/test.py
@ -136,7 +136,7 @@ def test_put(started_cluster, maybe_auth, positive, compression):
    values_csv = "1,2,3\n3,2,1\n78,43,45\n"
    filename = "test.csv"
    put_query = f"""insert into table function s3('http://{started_cluster.minio_ip}:{started_cluster.minio_port}/{bucket}/{filename}',
-                    {maybe_auth}'CSV', '{table_format}', {compression}) values {values}"""
+                    {maybe_auth}'CSV', '{table_format}', {compression}) values settings s3_truncate_on_insert=1 {values}"""

    try:
        run_query(instance, put_query)
@ -298,7 +298,7 @@ def test_put_csv(started_cluster, maybe_auth, positive):
    instance = started_cluster.instances["dummy"]  # type: ClickHouseInstance
    table_format = "column1 UInt32, column2 UInt32, column3 UInt32"
    filename = "test.csv"
-    put_query = "insert into table function s3('http://{}:{}/{}/{}', {}'CSV', '{}') format CSV".format(
+    put_query = "insert into table function s3('http://{}:{}/{}/{}', {}'CSV', '{}') format CSV settings s3_truncate_on_insert=1".format(
        started_cluster.minio_ip, MINIO_INTERNAL_PORT, bucket, filename, maybe_auth, table_format)
    csv_data = "8,9,16\n11,18,13\n22,14,2\n"

@ -322,7 +322,7 @@ def test_put_get_with_redirect(started_cluster):
    values = "(1, 1, 1), (1, 1, 1), (11, 11, 11)"
    values_csv = "1,1,1\n1,1,1\n11,11,11\n"
    filename = "test.csv"
-    query = "insert into table function s3('http://{}:{}/{}/{}', 'CSV', '{}') values {}".format(
+    query = "insert into table function s3('http://{}:{}/{}/{}', 'CSV', '{}') values settings s3_truncate_on_insert=1 {}".format(
        started_cluster.minio_redirect_host, started_cluster.minio_redirect_port, bucket, filename, table_format, values)
    run_query(instance, query)

@ -350,12 +350,12 @@ def test_put_with_zero_redirect(started_cluster):
    filename = "test.csv"

    # Should work without redirect
-    query = "insert into table function s3('http://{}:{}/{}/{}', 'CSV', '{}') values {}".format(
+    query = "insert into table function s3('http://{}:{}/{}/{}', 'CSV', '{}') values settings s3_truncate_on_insert=1 {}".format(
        started_cluster.minio_ip, MINIO_INTERNAL_PORT, bucket, filename, table_format, values)
    run_query(instance, query)

    # Should not work with redirect
-    query = "insert into table function s3('http://{}:{}/{}/{}', 'CSV', '{}') values {}".format(
+    query = "insert into table function s3('http://{}:{}/{}/{}', 'CSV', '{}') values settings s3_truncate_on_insert=1 {}".format(
        started_cluster.minio_redirect_host, started_cluster.minio_redirect_port, bucket, filename, table_format, values)
    exception_raised = False
    try:
@ -805,13 +805,13 @@ def test_seekable_formats(started_cluster):
    instance = started_cluster.instances["dummy"]  # type: ClickHouseInstance

    table_function = f"s3(s3_parquet, structure='a Int32, b String', format='Parquet')"
-    instance.query(f"insert into table function {table_function} SELECT number, randomString(100) FROM numbers(5000000)")
+    instance.query(f"insert into table function {table_function} SELECT number, randomString(100) FROM numbers(5000000) settings s3_truncate_on_insert=1")

    result = instance.query(f"SELECT count() FROM {table_function}")
    assert(int(result) == 5000000)

    table_function = f"s3(s3_orc, structure='a Int32, b String', format='ORC')"
-    exec_query_with_retry(instance, f"insert into table function {table_function} SELECT number, randomString(100) FROM numbers(5000000)")
+    exec_query_with_retry(instance, f"insert into table function {table_function} SELECT number, randomString(100) FROM numbers(5000000) settings s3_truncate_on_insert=1")

    result = instance.query(f"SELECT count() FROM {table_function}")
    assert(int(result) == 5000000)
@ -827,14 +827,14 @@ def test_seekable_formats_url(started_cluster):
    instance = started_cluster.instances["dummy"]

    table_function = f"s3(s3_parquet, structure='a Int32, b String', format='Parquet')"
-    instance.query(f"insert into table function {table_function} select number, randomString(100) from numbers(5000000)")
+    instance.query(f"insert into table function {table_function} select number, randomString(100) from numbers(5000000) settings s3_truncate_on_insert=1")

    table_function = f"url('http://{started_cluster.minio_host}:{started_cluster.minio_port}/{bucket}/test_parquet', 'Parquet', 'a Int32, b String')"
    result = instance.query(f"SELECT count() FROM {table_function}")
    assert(int(result) == 5000000)

    table_function = f"s3(s3_orc, structure='a Int32, b String', format='ORC')"
-    exec_query_with_retry(instance, f"insert into table function {table_function} select number, randomString(100) from numbers(5000000)")
+    exec_query_with_retry(instance, f"insert into table function {table_function} select number, randomString(100) from numbers(5000000) settings s3_truncate_on_insert=1")

    table_function = f"url('http://{started_cluster.minio_host}:{started_cluster.minio_port}/{bucket}/test_orc', 'ORC', 'a Int32, b String')"
    result = instance.query(f"SELECT count() FROM {table_function}")
@ -917,6 +917,48 @@ def test_empty_file(started_cluster):
    assert(int(result) == 0)


+def test_overwrite(started_cluster):
+    bucket = started_cluster.minio_bucket
+    instance = started_cluster.instances["dummy"]
+
+    table_function = f"s3(s3_parquet, structure='a Int32, b String', format='Parquet')"
+    instance.query(f"create table test_overwrite as {table_function}")
+    instance.query(f"truncate table test_overwrite")
+    instance.query(f"insert into test_overwrite select number, randomString(100) from numbers(50) settings s3_truncate_on_insert=1")
+    instance.query_and_get_error(f"insert into test_overwrite select number, randomString(100) from numbers(100)")
+    instance.query(f"insert into test_overwrite select number, randomString(100) from numbers(200) settings s3_truncate_on_insert=1")
+
+    result = instance.query(f"select count() from test_overwrite")
+    assert(int(result) == 200)
+
+
+def test_create_new_files_on_insert(started_cluster):
+    bucket = started_cluster.minio_bucket
+    instance = started_cluster.instances["dummy"]
+
+    table_function = f"s3(s3_parquet, structure='a Int32, b String', format='Parquet')"
+    instance.query(f"create table test_multiple_inserts as {table_function}")
+    instance.query(f"truncate table test_multiple_inserts")
+    instance.query(f"insert into test_multiple_inserts select number, randomString(100) from numbers(10) settings s3_truncate_on_insert=1")
+    instance.query(f"insert into test_multiple_inserts select number, randomString(100) from numbers(20) settings s3_create_new_file_on_insert=1")
+    instance.query(f"insert into test_multiple_inserts select number, randomString(100) from numbers(30) settings s3_create_new_file_on_insert=1")
+    
+    result = instance.query(f"select count() from test_multiple_inserts")
+    assert(int(result) == 60)
+
+    instance.query(f"drop table test_multiple_inserts")
+
+    table_function = f"s3(s3_parquet_gz, structure='a Int32, b String', format='Parquet')"
+    instance.query(f"create table test_multiple_inserts as {table_function}")
+    instance.query(f"truncate table test_multiple_inserts")
+    instance.query(f"insert into test_multiple_inserts select number, randomString(100) from numbers(10) settings s3_truncate_on_insert=1")
+    instance.query(f"insert into test_multiple_inserts select number, randomString(100) from numbers(20) settings s3_create_new_file_on_insert=1")
+    instance.query(f"insert into test_multiple_inserts select number, randomString(100) from numbers(30) settings s3_create_new_file_on_insert=1")
+    
+    result = instance.query(f"select count() from test_multiple_inserts")
+    assert(int(result) == 60)
+
+    
 def test_format_detection(started_cluster):
    bucket = started_cluster.minio_bucket
    instance = started_cluster.instances["dummy"]
--- a/tests/queries/0_stateless/01059_storage_file_compression.sh
+++ b/tests/queries/0_stateless/01059_storage_file_compression.sh
@ -12,7 +12,6 @@ do
    ${CLICKHOUSE_CLIENT} --query "CREATE TABLE file (x UInt64) ENGINE = File(TSV, '${CLICKHOUSE_DATABASE}/${m}.tsv.${m}')"
    ${CLICKHOUSE_CLIENT} --query "TRUNCATE TABLE file"
    ${CLICKHOUSE_CLIENT} --query "INSERT INTO file SELECT * FROM numbers(1000000)"
-    sleep 1
    ${CLICKHOUSE_CLIENT} --query "SELECT count(), max(x) FROM file"
    ${CLICKHOUSE_CLIENT} --query "DROP TABLE file"
 done
--- a/tests/queries/0_stateless/02125_lz4_compression_bug.reference
+++ b/tests/queries/0_stateless/02125_lz4_compression_bug.reference
@ -0,0 +1,45 @@
+Native
+9999
+99999
+999999
+2499999
+Values
+9999
+99999
+999999
+2499999
+JSONCompactEachRow
+9999
+99999
+999999
+2499999
+TSKV
+9999
+99999
+999999
+2499999
+TSV
+9999
+99999
+999999
+2499999
+CSV
+9999
+99999
+999999
+2499999
+JSONEachRow
+9999
+99999
+999999
+2499999
+JSONCompactEachRow
+9999
+99999
+999999
+2499999
+JSONStringsEachRow
+9999
+99999
+999999
+2499999
--- a/tests/queries/0_stateless/02125_lz4_compression_bug.sh
+++ b/tests/queries/0_stateless/02125_lz4_compression_bug.sh
@ -0,0 +1,21 @@
+#!/usr/bin/env bash
+# Tags: no-parallel
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+# shellcheck source=../shell_config.sh
+. "$CURDIR"/../shell_config.sh
+
+for format in Native Values JSONCompactEachRow TSKV TSV CSV JSONEachRow JSONCompactEachRow JSONStringsEachRow
+do
+    echo $format
+    ${CLICKHOUSE_CLIENT} --query "DROP TABLE IF EXISTS file"
+    ${CLICKHOUSE_CLIENT} --query "CREATE TABLE file (x UInt64) ENGINE = File($format, '${CLICKHOUSE_DATABASE}/data.$format.lz4')"
+    for size in 10000 100000 1000000 2500000
+    do
+        ${CLICKHOUSE_CLIENT} --query "TRUNCATE TABLE file"
+        ${CLICKHOUSE_CLIENT} --query "INSERT INTO file SELECT * FROM numbers($size)"
+        ${CLICKHOUSE_CLIENT} --query "SELECT max(x) FROM file"
+    done
+done
+
+${CLICKHOUSE_CLIENT} --query "DROP TABLE file"
--- a/tests/queries/0_stateless/02155_multiple_inserts_for_formats_with_suffix.reference
+++ b/tests/queries/0_stateless/02155_multiple_inserts_for_formats_with_suffix.reference
@ -0,0 +1,100 @@
+0
+1
+2
+3
+4
+5
+6
+7
+8
+9
+10
+11
+12
+13
+14
+15
+16
+17
+18
+19
+0
+1
+2
+3
+4
+5
+6
+7
+8
+9
+10
+11
+12
+13
+14
+15
+16
+17
+18
+19
+0
+1
+2
+3
+4
+5
+6
+7
+8
+9
+10
+11
+12
+13
+14
+15
+16
+17
+18
+19
+0
+1
+2
+3
+4
+5
+6
+7
+8
+9
+10
+11
+12
+13
+14
+15
+16
+17
+18
+19
+0
+1
+2
+3
+4
+5
+6
+7
+8
+9
+10
+11
+12
+13
+14
+15
+16
+17
+18
+19
--- a/tests/queries/0_stateless/02155_multiple_inserts_for_formats_with_suffix.sql
+++ b/tests/queries/0_stateless/02155_multiple_inserts_for_formats_with_suffix.sql
@ -0,0 +1,39 @@
+-- Tags: no-fasttest, no-parallel
+
+drop table if exists test;
+create table test (number UInt64) engine=File('Parquet');
+insert into test select * from numbers(10);
+insert into test select * from numbers(10, 10); -- { serverError CANNOT_APPEND_TO_FILE }
+insert into test select * from numbers(10, 10) settings engine_file_allow_create_multiple_files=1;
+select * from test order by number;
+truncate table test;
+drop table test;
+
+create table test (number UInt64) engine=File('Parquet', 'test_02155/test1/data.Parquet');
+insert into test select * from numbers(10) settings engine_file_truncate_on_insert=1;
+insert into test select * from numbers(10, 10); -- { serverError CANNOT_APPEND_TO_FILE }
+insert into test select * from numbers(10, 10) settings engine_file_allow_create_multiple_files=1;
+select * from test order by number;
+drop table test;
+
+
+insert into table function file(concat(currentDatabase(), '/test2/data.Parquet'), 'Parquet', 'number UInt64') select * from numbers(10) settings engine_file_truncate_on_insert=1;
+insert into table function file(concat(currentDatabase(), '/test2/data.Parquet'), 'Parquet', 'number UInt64') select * from numbers(10, 10); -- { serverError CANNOT_APPEND_TO_FILE }
+insert into table function file(concat(currentDatabase(), '/test2/data.Parquet'), 'Parquet', 'number UInt64') select * from numbers(10, 10) settings engine_file_allow_create_multiple_files=1;
+select * from file(concat(currentDatabase(), '/test2/data.Parquet'), 'Parquet', 'number UInt64');
+select * from file(concat(currentDatabase(), '/test2/data.1.Parquet'), 'Parquet', 'number UInt64');
+
+create table test (number UInt64) engine=File('Parquet', 'test_02155/test3/data.Parquet.gz');
+insert into test select * from numbers(10) settings engine_file_truncate_on_insert=1;
+;
+insert into test select * from numbers(10, 10); -- { serverError CANNOT_APPEND_TO_FILE }
+insert into test select * from numbers(10, 10) settings engine_file_allow_create_multiple_files=1;
+select * from test order by number;
+drop table test;
+
+insert into table function file(concat(currentDatabase(), '/test4/data.Parquet.gz'), 'Parquet', 'number UInt64') select * from numbers(10) settings engine_file_truncate_on_insert=1;
+insert into table function file(concat(currentDatabase(), '/test4/data.Parquet.gz'), 'Parquet', 'number UInt64') select * from numbers(10, 10); -- { serverError CANNOT_APPEND_TO_FILE }
+insert into table function file(concat(currentDatabase(), '/test4/data.Parquet.gz'), 'Parquet', 'number UInt64') select * from numbers(10, 10) settings engine_file_allow_create_multiple_files=1;
+select * from file(concat(currentDatabase(), '/test4/data.Parquet.gz'), 'Parquet', 'number UInt64');
+select * from file(concat(currentDatabase(), '/test4/data.1.Parquet.gz'), 'Parquet', 'number UInt64');
+
--- a/tests/queries/0_stateless/02168_avro_bug.sql
+++ b/tests/queries/0_stateless/02168_avro_bug.sql
@ -1,5 +1,5 @@
-- Tags: no-fasttest
-insert into table function file('data.avro', 'Avro', 'x UInt64') select * from numbers(10);
-insert into table function file('data.avro', 'Avro', 'x UInt64') select * from numbers(10);
-insert into table function file('data.avro', 'Avro', 'x UInt64') select * from numbers(10);
+-- Tags: no-fasttest, no-parallel
+insert into table function file('data.avro', 'Parquet', 'x UInt64') select * from numbers(10);
+insert into table function file('data.avro', 'Parquet', 'x UInt64') select * from numbers(10); -- { serverError CANNOT_APPEND_TO_FILE }
+insert into table function file('data.avro', 'Parquet', 'x UInt64') select * from numbers(10); -- { serverError CANNOT_APPEND_TO_FILE }
 select 'OK';
--- a/utils/CMakeLists.txt
+++ b/utils/CMakeLists.txt
@ -2,7 +2,7 @@ if (USE_CLANG_TIDY)
    set (CMAKE_CXX_CLANG_TIDY "${CLANG_TIDY_PATH}")
 endif ()

-if(MAKE_STATIC_LIBRARIES)
+if(USE_STATIC_LIBRARIES)
    set(MAX_LINKER_MEMORY 3500)
 else()
    set(MAX_LINKER_MEMORY 2500)
				`@ -1 +0,0 @@`
				`../../../en/faq/general/columnar-database.md`