ClickHouse/src/QueryPipeline/RemoteQueryExecutor.cpp

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

691 lines
25 KiB
C++
Raw Normal View History

2021-09-02 14:27:19 +00:00
#include <Common/ConcurrentBoundedQueue.h>
2021-10-15 20:18:20 +00:00
#include <QueryPipeline/RemoteQueryExecutor.h>
#include <QueryPipeline/RemoteQueryExecutorReadContext.h>
2020-06-02 15:59:57 +00:00
2020-06-02 16:27:05 +00:00
#include <Columns/ColumnConst.h>
#include <Common/CurrentThread.h>
2021-08-30 11:04:59 +00:00
#include "Core/Protocol.h"
2022-05-23 19:47:32 +00:00
#include <Processors/QueryPlan/BuildQueryPipelineSettings.h>
#include <Processors/QueryPlan/Optimizations/QueryPlanOptimizationSettings.h>
2020-06-02 16:27:05 +00:00
#include <Processors/Sources/SourceFromSingleChunk.h>
#include <Processors/Transforms/LimitsCheckingTransform.h>
2022-05-23 19:47:32 +00:00
#include <Processors/QueryPlan/QueryPlan.h>
#include <QueryPipeline/QueryPipelineBuilder.h>
2020-06-02 16:27:05 +00:00
#include <Storages/IStorage.h>
#include <Storages/SelectQueryInfo.h>
2020-06-02 16:27:05 +00:00
#include <Interpreters/castColumn.h>
#include <Interpreters/Cluster.h>
#include <Interpreters/Context.h>
2020-06-02 16:27:05 +00:00
#include <Interpreters/InternalTextLogsQueue.h>
Fix terribly broken, fragile and potentially cyclic linking Sorry for the clickbaity title. This is about static method ConnectionTimeouts::getHTTPTimeouts(). It was be declared in header IO/ConnectionTimeouts.h, and defined in header IO/ConnectionTimeoutsContext.h (!). This is weird and caused issues with linking on s390x (##45520). There was an attempt to fix some inconsistencies (#45848) but neither did @Algunenano nor me at first really understand why the definition is in the header. Turns out that ConnectionTimeoutsContext.h is only #include'd from source files which are part of the normal server build BUT NOT part of the keeper standalone build (which must be enabled via CMake -DBUILD_STANDALONE_KEEPER=1). This dependency was not documented and as a result, some misguided workarounds were introduced earlier, e.g. https://github.com/ClickHouse/ClickHouse/pull/38475/commits/0341c6c54bd7ac77200b4ca123208b195514ef20 The deeper cause was that getHTTPTimeouts() is passed a "Context". This class is part of the "dbms" libary which is deliberately not linked by the standalone build of clickhouse-keeper. The context is only used to read the settings and the "Settings" class is part of the clickhouse_common library which is linked by clickhouse-keeper already. To resolve this mess, this PR - creates source file IO/ConnectionTimeouts.cpp and moves all ConnectionTimeouts definitions into it, including getHTTPTimeouts(). - breaks the wrong dependency by passing "Settings" instead of "Context" into getHTTPTimeouts(). - resolves the previous hacks
2023-02-03 10:54:49 +00:00
#include <IO/ConnectionTimeouts.h>
2021-01-19 19:21:06 +00:00
#include <Client/MultiplexedConnections.h>
#include <Client/HedgedConnections.h>
#include <Storages/MergeTree/MergeTreeDataPartUUID.h>
#include <Storages/StorageMemory.h>
2020-06-02 16:27:05 +00:00
2021-08-31 23:47:52 +00:00
namespace ProfileEvents
{
extern const Event ReadTaskRequestsReceived;
extern const Event MergeTreeReadTaskRequestsReceived;
}
2020-06-02 16:27:05 +00:00
namespace DB
{
namespace ErrorCodes
{
2021-04-12 17:07:01 +00:00
extern const int LOGICAL_ERROR;
2020-06-02 16:27:05 +00:00
extern const int UNKNOWN_PACKET_FROM_SERVER;
extern const int DUPLICATED_PART_UUIDS;
2021-10-12 21:15:05 +00:00
extern const int SYSTEM_ERROR;
2020-06-02 16:27:05 +00:00
}
RemoteQueryExecutor::RemoteQueryExecutor(
const String & query_, const Block & header_, ContextPtr context_,
const Scalars & scalars_, const Tables & external_tables_,
QueryProcessingStage::Enum stage_, std::optional<Extension> extension_)
: header(header_), query(query_), context(context_), scalars(scalars_)
, external_tables(external_tables_), stage(stage_)
, task_iterator(extension_ ? extension_->task_iterator : nullptr)
, parallel_reading_coordinator(extension_ ? extension_->parallel_reading_coordinator : nullptr)
{}
2020-06-02 16:27:05 +00:00
RemoteQueryExecutor::RemoteQueryExecutor(
Connection & connection,
2021-04-08 14:22:19 +00:00
const String & query_, const Block & header_, ContextPtr context_,
ThrottlerPtr throttler, const Scalars & scalars_, const Tables & external_tables_,
QueryProcessingStage::Enum stage_, std::optional<Extension> extension_)
: RemoteQueryExecutor(query_, header_, context_, scalars_, external_tables_, stage_, extension_)
2020-06-02 16:27:05 +00:00
{
create_connections = [this, &connection, throttler, extension_]()
2020-06-02 16:27:05 +00:00
{
auto res = std::make_unique<MultiplexedConnections>(connection, context->getSettingsRef(), throttler);
if (extension_ && extension_->replica_info)
res->setReplicaInfo(*extension_->replica_info);
return res;
2020-06-02 16:27:05 +00:00
};
}
RemoteQueryExecutor::RemoteQueryExecutor(
std::shared_ptr<Connection> connection_ptr,
const String & query_, const Block & header_, ContextPtr context_,
ThrottlerPtr throttler, const Scalars & scalars_, const Tables & external_tables_,
QueryProcessingStage::Enum stage_, std::optional<Extension> extension_)
: RemoteQueryExecutor(query_, header_, context_, scalars_, external_tables_, stage_, extension_)
{
create_connections = [this, connection_ptr, throttler, extension_]()
{
auto res = std::make_unique<MultiplexedConnections>(connection_ptr, context->getSettingsRef(), throttler);
if (extension_ && extension_->replica_info)
res->setReplicaInfo(*extension_->replica_info);
return res;
2020-06-02 16:27:05 +00:00
};
}
RemoteQueryExecutor::RemoteQueryExecutor(
2021-01-19 19:21:06 +00:00
std::vector<IConnectionPool::Entry> && connections_,
2021-04-08 14:22:19 +00:00
const String & query_, const Block & header_, ContextPtr context_,
const ThrottlerPtr & throttler, const Scalars & scalars_, const Tables & external_tables_,
QueryProcessingStage::Enum stage_, std::optional<Extension> extension_)
2021-04-08 14:22:19 +00:00
: header(header_), query(query_), context(context_)
, scalars(scalars_), external_tables(external_tables_), stage(stage_)
, task_iterator(extension_ ? extension_->task_iterator : nullptr)
, parallel_reading_coordinator(extension_ ? extension_->parallel_reading_coordinator : nullptr)
2020-06-02 16:27:05 +00:00
{
create_connections = [this, connections_, throttler, extension_]() mutable {
auto res = std::make_unique<MultiplexedConnections>(std::move(connections_), context->getSettingsRef(), throttler);
if (extension_ && extension_->replica_info)
res->setReplicaInfo(*extension_->replica_info);
return res;
2020-06-02 16:27:05 +00:00
};
}
RemoteQueryExecutor::RemoteQueryExecutor(
const ConnectionPoolWithFailoverPtr & pool,
2021-04-08 14:22:19 +00:00
const String & query_, const Block & header_, ContextPtr context_,
const ThrottlerPtr & throttler, const Scalars & scalars_, const Tables & external_tables_,
QueryProcessingStage::Enum stage_, std::optional<Extension> extension_)
2021-04-08 14:22:19 +00:00
: header(header_), query(query_), context(context_)
, scalars(scalars_), external_tables(external_tables_), stage(stage_)
, task_iterator(extension_ ? extension_->task_iterator : nullptr)
, parallel_reading_coordinator(extension_ ? extension_->parallel_reading_coordinator : nullptr)
2020-06-02 16:27:05 +00:00
{
create_connections = [this, pool, throttler, extension_]()->std::unique_ptr<IConnections>
2020-06-02 16:27:05 +00:00
{
const Settings & current_settings = context->getSettingsRef();
2020-06-02 16:27:05 +00:00
auto timeouts = ConnectionTimeouts::getTCPTimeoutsWithFailover(current_settings);
2021-01-19 19:21:06 +00:00
2021-02-01 17:23:46 +00:00
#if defined(OS_LINUX)
if (current_settings.use_hedged_requests)
2020-06-02 16:27:05 +00:00
{
2021-01-19 19:21:06 +00:00
std::shared_ptr<QualifiedTableName> table_to_check = nullptr;
if (main_table)
table_to_check = std::make_shared<QualifiedTableName>(main_table.getQualifiedName());
auto res = std::make_unique<HedgedConnections>(pool, context, timeouts, throttler, pool_mode, table_to_check);
if (extension_ && extension_->replica_info)
res->setReplicaInfo(*extension_->replica_info);
return res;
2020-06-02 16:27:05 +00:00
}
2021-02-01 17:23:46 +00:00
#endif
2020-06-02 16:27:05 +00:00
2021-02-01 17:23:46 +00:00
std::vector<IConnectionPool::Entry> connection_entries;
if (main_table)
{
auto try_results = pool->getManyChecked(timeouts, &current_settings, pool_mode, main_table.getQualifiedName());
connection_entries.reserve(try_results.size());
for (auto & try_result : try_results)
connection_entries.emplace_back(std::move(try_result.entry));
2021-01-19 19:21:06 +00:00
}
2021-02-01 17:23:46 +00:00
else
connection_entries = pool->getMany(timeouts, &current_settings, pool_mode);
auto res = std::make_unique<MultiplexedConnections>(std::move(connection_entries), current_settings, throttler);
if (extension_ && extension_->replica_info)
res->setReplicaInfo(*extension_->replica_info);
return res;
2020-06-02 16:27:05 +00:00
};
}
RemoteQueryExecutor::~RemoteQueryExecutor()
{
/** If interrupted in the middle of the loop of communication with replicas, then interrupt
* all connections, then read and skip the remaining packets to make sure
* these connections did not remain hanging in the out-of-sync state.
*/
if (established || isQueryPending())
2021-01-19 19:21:06 +00:00
connections->disconnect();
2020-06-02 16:27:05 +00:00
}
/** If we receive a block with slightly different column types, or with excessive columns,
* we will adapt it to expected structure.
*/
2021-03-24 18:36:31 +00:00
static Block adaptBlockStructure(const Block & block, const Block & header)
2020-06-02 16:27:05 +00:00
{
/// Special case when reader doesn't care about result structure. Deprecated and used only in Benchmark, PerformanceTest.
if (!header)
return block;
Block res;
res.info = block.info;
for (const auto & elem : header)
{
ColumnPtr column;
if (elem.column && isColumnConst(*elem.column))
{
/// We expect constant column in block.
/// If block is not empty, then get value for constant from it,
/// because it may be different for remote server for functions like version(), uptime(), ...
if (block.rows() > 0 && block.has(elem.name))
{
/// Const column is passed as materialized. Get first value from it.
///
/// TODO: check that column contains the same value.
/// TODO: serialize const columns.
auto col = block.getByName(elem.name);
col.column = block.getByName(elem.name).column->cut(0, 1);
column = castColumn(col, elem.type);
if (!isColumnConst(*column))
column = ColumnConst::create(column, block.rows());
else
/// It is not possible now. Just in case we support const columns serialization.
column = column->cloneResized(block.rows());
}
else
column = elem.column->cloneResized(block.rows());
}
else
column = castColumn(block.getByName(elem.name), elem.type);
res.insert({column, elem.type, elem.name});
}
return res;
}
Fix `parallel_reading_from_replicas` with `clickhouse-bechmark` (#34751) * Use INITIAL_QUERY for clickhouse-benchmark Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com> * Fix parallel_reading_from_replicas with clickhouse-bechmark Before it produces the following error: $ clickhouse-benchmark --stacktrace -i1 --query "select * from remote('127.1', default.data_mt) limit 10" --allow_experimental_parallel_reading_from_replicas=1 --max_parallel_replicas=3 Loaded 1 queries. Logical error: 'Coordinator for parallel reading from replicas is not initialized'. Aborted (core dumped) Since it uses the same code, i.e RemoteQueryExecutor -> MultiplexedConnections, which enables coordinator if it was requested from settings, but it should be done only for non-initial queries, i.e. when server send connection to another server. Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com> * Fix 02226_parallel_reading_from_replicas_benchmark for older shellcheck By shellcheck 0.8 does not complains, while on CI shellcheck 0.7.0 and it does complains [1]: In 02226_parallel_reading_from_replicas_benchmark.sh line 17: --allow_experimental_parallel_reading_from_replicas=1 ^-- SC2191: The = here is literal. To assign by index, use ( [index]=value ) with no spaces. To keep as literal, quote it. Did you mean: "--allow_experimental_parallel_reading_from_replicas=1" [1]: https://s3.amazonaws.com/clickhouse-test-reports/34751/d883af711822faf294c876b017cbf745b1cda1b3/style_check__actions_/shellcheck_output.txt Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com>
2022-03-08 15:42:29 +00:00
void RemoteQueryExecutor::sendQuery(ClientInfo::QueryKind query_kind)
2020-06-02 16:27:05 +00:00
{
if (sent_query)
return;
2021-01-19 19:21:06 +00:00
connections = create_connections();
2020-06-02 16:27:05 +00:00
const auto & settings = context->getSettingsRef();
2021-01-19 19:21:06 +00:00
if (settings.skip_unavailable_shards && 0 == connections->size())
2020-06-02 16:27:05 +00:00
return;
Fix "Unexpected packet Data received from client" error Fix query cancelation in case of Distributed queries with LIMIT (when the initator does not required to read all the data), since this cannot be done until the query was sent (from the Query packet up to the empty data Block), otherwise you will get: 2020.11.21 21:47:23.297161 [ 184 ] {} <Error> TCPHandler: Code: 101, e.displayText() = DB::Exception: Unexpected packet Data received from client, Stack trace: 0. /build/obj-x86_64-linux-gnu/../contrib/libcxx/include/exception:129: Poco::Exception::Exception(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, int) @ 0x244f5bc9 in /usr/bin/clickhouse 1. /build/obj-x86_64-linux-gnu/../src/Common/Exception.cpp:40: DB::Exception::Exception(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, int) @ 0xa14a421 in /usr/bin/clickhouse 2. /build/obj-x86_64-linux-gnu/../src/Common/NetException.h:0: DB::TCPHandler::receiveUnexpectedData() @ 0x1e032a74 in /usr/bin/clickhouse 3. /build/obj-x86_64-linux-gnu/../src/Server/TCPHandler.cpp:824: DB::TCPHandler::receivePacket() @ 0x1e024685 in /usr/bin/clickhouse 4. /build/obj-x86_64-linux-gnu/../src/Server/TCPHandler.cpp:178: DB::TCPHandler::runImpl() @ 0x1e01736b in /usr/bin/clickhouse 5. /build/obj-x86_64-linux-gnu/../src/Server/TCPHandler.cpp:0: DB::TCPHandler::run() @ 0x1e035c1b in /usr/bin/clickhouse 6. /build/obj-x86_64-linux-gnu/../contrib/poco/Net/src/TCPServerConnection.cpp:57: Poco::Net::TCPServerConnection::start() @ 0x243559cf in /usr/bin/clickhouse 7. /build/obj-x86_64-linux-gnu/../contrib/poco/Net/src/TCPServerDispatcher.cpp:0: Poco::Net::TCPServerDispatcher::run() @ 0x24356521 in /usr/bin/clickhouse 8. /build/obj-x86_64-linux-gnu/../contrib/poco/Foundation/src/ThreadPool.cpp:0: Poco::PooledThread::run() @ 0x24609175 in /usr/bin/clickhouse 9. /build/obj-x86_64-linux-gnu/../contrib/poco/Foundation/src/Thread_POSIX.cpp:0: Poco::ThreadImpl::runnableEntry(void*) @ 0x24603cb7 in /usr/bin/clickhouse 10. start_thread @ 0x9669 in /usr/lib/x86_64-linux-gnu/libpthread-2.30.so 11. __clone @ 0x1222b3 in /usr/lib/x86_64-linux-gnu/libc-2.30.so
2020-11-21 21:20:00 +00:00
/// Query cannot be canceled in the middle of the send query,
2020-11-28 05:37:54 +00:00
/// since there are multiple packets:
Fix "Unexpected packet Data received from client" error Fix query cancelation in case of Distributed queries with LIMIT (when the initator does not required to read all the data), since this cannot be done until the query was sent (from the Query packet up to the empty data Block), otherwise you will get: 2020.11.21 21:47:23.297161 [ 184 ] {} <Error> TCPHandler: Code: 101, e.displayText() = DB::Exception: Unexpected packet Data received from client, Stack trace: 0. /build/obj-x86_64-linux-gnu/../contrib/libcxx/include/exception:129: Poco::Exception::Exception(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, int) @ 0x244f5bc9 in /usr/bin/clickhouse 1. /build/obj-x86_64-linux-gnu/../src/Common/Exception.cpp:40: DB::Exception::Exception(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, int) @ 0xa14a421 in /usr/bin/clickhouse 2. /build/obj-x86_64-linux-gnu/../src/Common/NetException.h:0: DB::TCPHandler::receiveUnexpectedData() @ 0x1e032a74 in /usr/bin/clickhouse 3. /build/obj-x86_64-linux-gnu/../src/Server/TCPHandler.cpp:824: DB::TCPHandler::receivePacket() @ 0x1e024685 in /usr/bin/clickhouse 4. /build/obj-x86_64-linux-gnu/../src/Server/TCPHandler.cpp:178: DB::TCPHandler::runImpl() @ 0x1e01736b in /usr/bin/clickhouse 5. /build/obj-x86_64-linux-gnu/../src/Server/TCPHandler.cpp:0: DB::TCPHandler::run() @ 0x1e035c1b in /usr/bin/clickhouse 6. /build/obj-x86_64-linux-gnu/../contrib/poco/Net/src/TCPServerConnection.cpp:57: Poco::Net::TCPServerConnection::start() @ 0x243559cf in /usr/bin/clickhouse 7. /build/obj-x86_64-linux-gnu/../contrib/poco/Net/src/TCPServerDispatcher.cpp:0: Poco::Net::TCPServerDispatcher::run() @ 0x24356521 in /usr/bin/clickhouse 8. /build/obj-x86_64-linux-gnu/../contrib/poco/Foundation/src/ThreadPool.cpp:0: Poco::PooledThread::run() @ 0x24609175 in /usr/bin/clickhouse 9. /build/obj-x86_64-linux-gnu/../contrib/poco/Foundation/src/Thread_POSIX.cpp:0: Poco::ThreadImpl::runnableEntry(void*) @ 0x24603cb7 in /usr/bin/clickhouse 10. start_thread @ 0x9669 in /usr/lib/x86_64-linux-gnu/libpthread-2.30.so 11. __clone @ 0x1222b3 in /usr/lib/x86_64-linux-gnu/libc-2.30.so
2020-11-21 21:20:00 +00:00
/// - Query
/// - Data (multiple times)
///
/// And after the Cancel packet none Data packet can be sent, otherwise the remote side will throw:
///
/// Unexpected packet Data received from client
///
std::lock_guard guard(was_cancelled_mutex);
2020-06-02 16:27:05 +00:00
established = true;
was_cancelled = false;
2020-06-02 16:27:05 +00:00
auto timeouts = ConnectionTimeouts::getTCPTimeoutsWithFailover(settings);
ClientInfo modified_client_info = context->getClientInfo();
Fix `parallel_reading_from_replicas` with `clickhouse-bechmark` (#34751) * Use INITIAL_QUERY for clickhouse-benchmark Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com> * Fix parallel_reading_from_replicas with clickhouse-bechmark Before it produces the following error: $ clickhouse-benchmark --stacktrace -i1 --query "select * from remote('127.1', default.data_mt) limit 10" --allow_experimental_parallel_reading_from_replicas=1 --max_parallel_replicas=3 Loaded 1 queries. Logical error: 'Coordinator for parallel reading from replicas is not initialized'. Aborted (core dumped) Since it uses the same code, i.e RemoteQueryExecutor -> MultiplexedConnections, which enables coordinator if it was requested from settings, but it should be done only for non-initial queries, i.e. when server send connection to another server. Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com> * Fix 02226_parallel_reading_from_replicas_benchmark for older shellcheck By shellcheck 0.8 does not complains, while on CI shellcheck 0.7.0 and it does complains [1]: In 02226_parallel_reading_from_replicas_benchmark.sh line 17: --allow_experimental_parallel_reading_from_replicas=1 ^-- SC2191: The = here is literal. To assign by index, use ( [index]=value ) with no spaces. To keep as literal, quote it. Did you mean: "--allow_experimental_parallel_reading_from_replicas=1" [1]: https://s3.amazonaws.com/clickhouse-test-reports/34751/d883af711822faf294c876b017cbf745b1cda1b3/style_check__actions_/shellcheck_output.txt Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com>
2022-03-08 15:42:29 +00:00
modified_client_info.query_kind = query_kind;
2020-06-02 16:27:05 +00:00
if (!duplicated_part_uuids.empty())
connections->sendIgnoredPartUUIDs(duplicated_part_uuids);
2021-01-19 19:21:06 +00:00
connections->sendQuery(timeouts, query, query_id, stage, modified_client_info, true);
2020-06-02 16:27:05 +00:00
established = false;
sent_query = true;
if (settings.enable_scalar_subquery_optimization)
sendScalars();
sendExternalTables();
}
2023-02-03 13:34:18 +00:00
Block RemoteQueryExecutor::readBlock()
{
while (true)
{
auto res = read();
if (res.getType() == ReadResult::Type::Data)
return res.getBlock();
}
}
RemoteQueryExecutor::ReadResult RemoteQueryExecutor::read()
2020-06-02 16:27:05 +00:00
{
if (!sent_query)
{
sendQuery();
if (context->getSettingsRef().skip_unavailable_shards && (0 == connections->size()))
2023-02-03 13:34:18 +00:00
return ReadResult(Block());
2020-06-02 16:27:05 +00:00
}
while (true)
{
Fix possible "No more packets are available" for distributed queries CI founds the following case: <details> 2022.05.25 22:36:06.778808 [ 3037 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Fatal> : Logical error: 'No more packets are available.'. 2022.05.25 22:42:24.960075 [ 17397 ] {} <Fatal> BaseDaemon: ######################################## 2022.05.25 22:42:24.971173 [ 17397 ] {} <Fatal> BaseDaemon: (version 22.6.1.1, build id: 9A1F9489854CED36) (from thread 3037) (query_id: 77743723-1fcd-4b3d-babc-d0615e3ff40e) (query: SELECT * FROM 2022.05.25 22:42:25.046871 [ 17397 ] {} <Fatal> BaseDaemon: 5. ./build_docker/../src/Common/Exception.cpp:47: DB::abortOnFailedAssertion() 2022.05.25 22:42:25.181449 [ 17397 ] {} <Fatal> BaseDaemon: 6. ./build_docker/../src/Common/Exception.cpp:70: DB::Exception::Exception() 2022.05.25 22:42:25.367710 [ 17397 ] {} <Fatal> BaseDaemon: 7. ./build_docker/../src/Client/MultiplexedConnections.cpp:0: DB::MultiplexedConnections::receivePacketUnlocked() 2022.05.25 22:42:25.414201 [ 17397 ] {} <Fatal> BaseDaemon: 8. ./build_docker/../src/Client/MultiplexedConnections.cpp:0: DB::MultiplexedConnections::receivePacket() 2022.05.25 22:42:25.493066 [ 17397 ] {} <Fatal> BaseDaemon: 9. ./build_docker/../src/QueryPipeline/RemoteQueryExecutor.cpp:279: DB::RemoteQueryExecutor::read() 2022.05.25 22:42:25.612679 [ 17397 ] {} <Fatal> BaseDaemon: 10. ./build_docker/../src/Processors/Sources/RemoteSource.cpp:0: DB::RemoteSource::tryGenerate() Here are additional logs for this query: $ pigz -cd clickhouse-server.stress.log.gz | fgrep -a 77743723-1fcd-4b3d-babc-d0615e3ff40e | fgrep -e Connection -e Distributed -e Fatal 2022.05.25 22:36:04.698671 [ 6613 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Trace> Connection (127.0.0.2:9000): Connecting. Database: (not specified). User: default 2022.05.25 22:36:04.722568 [ 3419 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Trace> Connection (127.0.0.2:9000): Connecting. Database: (not specified). User: default 2022.05.25 22:36:05.014432 [ 6613 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Trace> Connection (127.0.0.2:9000): Connected to ClickHouse server version 22.6.1. 2022.05.25 22:36:05.091397 [ 6613 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Debug> Connection (127.0.0.2:9000): Sent data for 2 scalars, total 2 rows in 0.000125814 sec., 15602 rows/sec., 68.00 B (517.81 KiB/sec.), compressed 0.4594594594594595 times to 148.00 B (1.10 MiB/sec.) 2022.05.25 22:36:05.301301 [ 3419 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Trace> Connection (127.0.0.2:9000): Connected to ClickHouse server version 22.6.1. 2022.05.25 22:36:05.343140 [ 3419 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Debug> Connection (127.0.0.2:9000): Sent data for 2 scalars, total 2 rows in 0.000116304 sec., 16889 rows/sec., 68.00 B (559.80 KiB/sec.), compressed 0.4594594594594595 times to 148.00 B (1.19 MiB/sec.) 2022.05.25 22:36:06.682535 [ 6613 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Trace> StorageDistributed (remote): (127.0.0.2:9000) Cancelling query because enough data has been read 2022.05.25 22:36:06.778808 [ 3037 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Fatal> : Logical error: 'No more packets are available.'. 2022.05.25 22:36:06.789505 [ 3419 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Trace> StorageDistributed (remote): (127.0.0.2:9000) Cancelling query because enough data has been read 2022.05.25 22:42:24.971173 [ 17397 ] {} <Fatal> BaseDaemon: (version 22.6.1.1, build id: 9A1F9489854CED36) (from thread 3037) (query_id: 77743723-1fcd-4b3d-babc-d0615e3ff40e) (query: SELECT * FROM </details> So between cancelling different sources the LOGICAL_ERROR occured, I believe that this is because of the race: T1: T2: RemoteQueryExecutor::read() checks was_cancelled RemoteQueryExecutor::tryCancel() connections->cancel() calls connections->receivePacket() Note, for this problem async_socket_for_remote/use_hedged_requests should be disabled, and original settings was: - --max_parallel_replicas=3 - --use_hedged_requests=false - --allow_experimental_parallel_reading_from_replicas=3 CI: https://s3.amazonaws.com/clickhouse-test-reports/37469/41cb029ed23e77f3a108e07e6b1b1bcb03dc7fcf/stress_test__undefined__actions_/fatal_messages.txt Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com>
2022-06-03 11:49:22 +00:00
std::lock_guard lock(was_cancelled_mutex);
2020-06-02 16:27:05 +00:00
if (was_cancelled)
2023-02-03 13:34:18 +00:00
return ReadResult(Block());
2020-06-02 16:27:05 +00:00
2023-02-03 13:34:18 +00:00
auto packet = connections->receivePacket();
auto anything = processPacket(std::move(packet));
2020-06-02 16:27:05 +00:00
2023-02-03 13:34:18 +00:00
if (anything.getType() == ReadResult::Type::Data || anything.getType() == ReadResult::Type::ParallelReplicasToken)
return anything;
if (got_duplicated_part_uuids)
return restartQueryWithoutDuplicatedUUIDs();
2020-12-02 17:02:14 +00:00
}
}
2023-02-03 13:34:18 +00:00
RemoteQueryExecutor::ReadResult RemoteQueryExecutor::read(std::unique_ptr<ReadContext> & read_context [[maybe_unused]])
2020-12-02 17:02:14 +00:00
{
2020-12-18 13:15:03 +00:00
#if defined(OS_LINUX)
2020-12-02 17:02:14 +00:00
if (!sent_query)
{
sendQuery();
if (context->getSettingsRef().skip_unavailable_shards && (0 == connections->size()))
2023-02-03 13:34:18 +00:00
return ReadResult(Block());
2020-12-02 17:02:14 +00:00
}
if (!read_context || resent_query)
2020-12-16 20:27:31 +00:00
{
std::lock_guard lock(was_cancelled_mutex);
2020-12-17 10:07:28 +00:00
if (was_cancelled)
2023-02-03 13:34:18 +00:00
return ReadResult(Block());
2020-12-17 10:07:28 +00:00
2021-01-19 19:21:06 +00:00
read_context = std::make_unique<ReadContext>(*connections);
2020-12-16 20:27:31 +00:00
}
2020-12-02 17:02:14 +00:00
2023-02-28 14:43:49 +00:00
while (true)
2020-12-03 12:21:10 +00:00
{
2020-12-14 16:18:12 +00:00
if (!read_context->resumeRoutine())
2023-02-03 13:34:18 +00:00
return ReadResult(Block());
2020-12-02 17:02:14 +00:00
if (read_context->is_read_in_progress.load(std::memory_order_relaxed))
2020-12-02 17:02:14 +00:00
{
2020-12-03 12:21:10 +00:00
read_context->setTimer();
2023-02-03 13:34:18 +00:00
return ReadResult(read_context->epoll.getFileDescriptor());
2020-12-02 17:02:14 +00:00
}
else
{
/// We need to check that query was not cancelled again,
/// to avoid the race between cancel() thread and read() thread.
/// (since cancel() thread will steal the fiber and may update the packet).
if (was_cancelled)
2023-02-03 13:34:18 +00:00
return ReadResult(Block());
2023-02-03 13:34:18 +00:00
auto anything = processPacket(std::move(read_context->packet));
if (anything.getType() == ReadResult::Type::Data || anything.getType() == ReadResult::Type::ParallelReplicasToken)
return anything;
if (got_duplicated_part_uuids)
2021-02-05 09:54:34 +00:00
return restartQueryWithoutDuplicatedUUIDs(&read_context);
2020-06-02 16:27:05 +00:00
}
}
2020-12-18 13:15:03 +00:00
#else
return read();
2020-12-14 16:16:08 +00:00
#endif
2020-12-18 13:15:03 +00:00
}
2020-12-02 17:02:14 +00:00
2021-02-05 09:54:34 +00:00
2023-02-03 13:34:18 +00:00
RemoteQueryExecutor::ReadResult RemoteQueryExecutor::restartQueryWithoutDuplicatedUUIDs(std::unique_ptr<ReadContext> * read_context)
2021-02-05 09:54:34 +00:00
{
/// Cancel previous query and disconnect before retry.
cancel(read_context);
connections->disconnect();
2021-02-05 09:54:34 +00:00
/// Only resend once, otherwise throw an exception
if (!resent_query)
{
if (log)
LOG_DEBUG(log, "Found duplicate UUIDs, will retry query without those parts");
resent_query = true;
sent_query = false;
got_duplicated_part_uuids = false;
/// Consecutive read will implicitly send query first.
if (!read_context)
return read();
else
return read(*read_context);
}
throw Exception(ErrorCodes::DUPLICATED_PART_UUIDS, "Found duplicate uuids while processing query");
2021-02-05 09:54:34 +00:00
}
2023-02-03 13:34:18 +00:00
RemoteQueryExecutor::ReadResult RemoteQueryExecutor::processPacket(Packet packet)
2020-12-02 17:02:14 +00:00
{
switch (packet.type)
{
case Protocol::Server::MergeTreeReadTaskRequest:
processMergeTreeReadTaskRequest(packet.request);
2023-02-03 13:34:18 +00:00
return ReadResult(ReadResult::Type::ParallelReplicasToken);
case Protocol::Server::MergeTreeAllRangesAnnounecement:
processMergeTreeInitialReadAnnounecement(packet.announcement);
return ReadResult(ReadResult::Type::ParallelReplicasToken);
case Protocol::Server::ReadTaskRequest:
2021-04-10 02:21:18 +00:00
processReadTaskRequest();
break;
case Protocol::Server::PartUUIDs:
if (!setPartUUIDs(packet.part_uuids))
got_duplicated_part_uuids = true;
break;
2020-12-02 17:02:14 +00:00
case Protocol::Server::Data:
2023-01-02 12:15:31 +00:00
/// Note: `packet.block.rows() > 0` means it's a header block.
/// We can actually return it, and the first call to RemoteQueryExecutor::read
/// will return earlier. We should consider doing it.
if (packet.block && (packet.block.rows() > 0))
2023-02-03 13:34:18 +00:00
return ReadResult(adaptBlockStructure(packet.block, header));
2020-12-02 17:02:14 +00:00
break; /// If the block is empty - we will receive other packets before EndOfStream.
case Protocol::Server::Exception:
got_exception_from_replica = true;
packet.exception->rethrow();
break;
case Protocol::Server::EndOfStream:
2021-01-19 19:21:06 +00:00
if (!connections->hasActiveConnections())
2020-12-02 17:02:14 +00:00
{
finished = true;
2023-02-03 13:34:18 +00:00
/// TODO: Replace with Type::Finished
return ReadResult(Block{});
2020-12-02 17:02:14 +00:00
}
break;
case Protocol::Server::Progress:
/** We use the progress from a remote server.
* We also include in ProcessList,
* and we use it to check
* constraints (for example, the minimum speed of query execution)
* and quotas (for example, the number of lines to read).
*/
if (progress_callback)
progress_callback(packet.progress);
break;
case Protocol::Server::ProfileInfo:
/// Use own (client-side) info about read bytes, it is more correct info than server-side one.
if (profile_info_callback)
profile_info_callback(packet.profile_info);
break;
case Protocol::Server::Totals:
totals = packet.block;
if (totals)
totals = adaptBlockStructure(totals, header);
2020-12-02 17:02:14 +00:00
break;
case Protocol::Server::Extremes:
extremes = packet.block;
if (extremes)
extremes = adaptBlockStructure(packet.block, header);
2020-12-02 17:02:14 +00:00
break;
case Protocol::Server::Log:
/// Pass logs from remote server to client
if (auto log_queue = CurrentThread::getInternalTextLogsQueue())
log_queue->pushBlock(std::move(packet.block));
break;
2021-08-30 11:04:59 +00:00
case Protocol::Server::ProfileEvents:
2021-09-01 14:47:12 +00:00
/// Pass profile events from remote server to client
if (auto profile_queue = CurrentThread::getInternalProfileEventsQueue())
2021-10-12 21:15:05 +00:00
if (!profile_queue->emplace(std::move(packet.block)))
throw Exception(ErrorCodes::SYSTEM_ERROR, "Could not push into profile queue");
break;
2021-08-30 11:04:59 +00:00
case Protocol::Server::TimezoneUpdate:
break;
2020-12-02 17:02:14 +00:00
default:
got_unknown_packet_from_replica = true;
throw Exception(
ErrorCodes::UNKNOWN_PACKET_FROM_SERVER,
"Unknown packet {} from one of the following replicas: {}",
packet.type,
2021-01-19 19:21:06 +00:00
connections->dumpAddresses());
2020-12-02 17:02:14 +00:00
}
2023-02-03 13:34:18 +00:00
return ReadResult(ReadResult::Type::Nothing);
2020-06-02 16:27:05 +00:00
}
bool RemoteQueryExecutor::setPartUUIDs(const std::vector<UUID> & uuids)
{
auto query_context = context->getQueryContext();
auto duplicates = query_context->getPartUUIDs()->add(uuids);
if (!duplicates.empty())
{
duplicated_part_uuids.insert(duplicated_part_uuids.begin(), duplicates.begin(), duplicates.end());
return false;
}
return true;
}
2021-04-10 02:21:18 +00:00
void RemoteQueryExecutor::processReadTaskRequest()
{
2021-04-10 02:21:18 +00:00
if (!task_iterator)
throw Exception(ErrorCodes::LOGICAL_ERROR, "Distributed task iterator is not initialized");
ProfileEvents::increment(ProfileEvents::ReadTaskRequestsReceived);
2021-04-10 02:21:18 +00:00
auto response = (*task_iterator)();
2021-04-08 19:00:39 +00:00
connections->sendReadTaskResponse(response);
}
2023-02-03 13:34:18 +00:00
void RemoteQueryExecutor::processMergeTreeReadTaskRequest(ParallelReadRequest request)
{
if (!parallel_reading_coordinator)
throw Exception(ErrorCodes::LOGICAL_ERROR, "Coordinator for parallel reading from replicas is not initialized");
ProfileEvents::increment(ProfileEvents::MergeTreeReadTaskRequestsReceived);
auto response = parallel_reading_coordinator->handleRequest(std::move(request));
connections->sendMergeTreeReadTaskResponse(response);
}
2023-02-03 13:34:18 +00:00
void RemoteQueryExecutor::processMergeTreeInitialReadAnnounecement(InitialAllRangesAnnouncement announcement)
{
if (!parallel_reading_coordinator)
throw Exception(ErrorCodes::LOGICAL_ERROR, "Coordinator for parallel reading from replicas is not initialized");
parallel_reading_coordinator->handleInitialAllRangesAnnouncement(announcement);
}
2020-12-07 13:47:11 +00:00
void RemoteQueryExecutor::finish(std::unique_ptr<ReadContext> * read_context)
2020-06-02 16:27:05 +00:00
{
/** If one of:
* - nothing started to do;
* - received all packets before EndOfStream;
* - received exception from one replica;
* - received an unknown packet from one replica;
* then you do not need to read anything.
*/
if (!isQueryPending() || hasThrownException())
return;
/** If you have not read all the data yet, but they are no longer needed.
* This may be due to the fact that the data is sufficient (for example, when using LIMIT).
*/
/// Send the request to abort the execution of the request, if not already sent.
2020-12-17 10:07:28 +00:00
tryCancel("Cancelling query because enough data has been read", read_context);
/// Get the remaining packets so that there is no out of sync in the connections to the replicas.
Packet packet = connections->drain();
switch (packet.type)
2020-06-02 16:27:05 +00:00
{
case Protocol::Server::EndOfStream:
finished = true;
break;
case Protocol::Server::Log:
/// Pass logs from remote server to client
if (auto log_queue = CurrentThread::getInternalTextLogsQueue())
log_queue->pushBlock(std::move(packet.block));
break;
case Protocol::Server::Exception:
got_exception_from_replica = true;
packet.exception->rethrow();
break;
case Protocol::Server::ProfileEvents:
/// Pass profile events from remote server to client
if (auto profile_queue = CurrentThread::getInternalProfileEventsQueue())
if (!profile_queue->emplace(std::move(packet.block)))
throw Exception(ErrorCodes::SYSTEM_ERROR, "Could not push into profile queue");
break;
case Protocol::Server::TimezoneUpdate:
break;
default:
got_unknown_packet_from_replica = true;
throw Exception(ErrorCodes::UNKNOWN_PACKET_FROM_SERVER, "Unknown packet {} from one of the following replicas: {}",
toString(packet.type),
connections->dumpAddresses());
2020-06-02 16:27:05 +00:00
}
}
2020-12-04 13:35:24 +00:00
void RemoteQueryExecutor::cancel(std::unique_ptr<ReadContext> * read_context)
2020-06-02 16:27:05 +00:00
{
{
std::lock_guard lock(external_tables_mutex);
/// Stop sending external data.
for (auto & vec : external_tables_data)
for (auto & elem : vec)
elem->is_cancelled = true;
}
if (!isQueryPending() || hasThrownException())
return;
2020-12-17 10:07:28 +00:00
tryCancel("Cancelling query", read_context);
2020-06-02 16:27:05 +00:00
}
void RemoteQueryExecutor::sendScalars()
{
2021-01-19 19:21:06 +00:00
connections->sendScalarsData(scalars);
2020-06-02 16:27:05 +00:00
}
void RemoteQueryExecutor::sendExternalTables()
{
2021-01-19 19:21:06 +00:00
size_t count = connections->size();
2020-06-02 16:27:05 +00:00
{
std::lock_guard lock(external_tables_mutex);
external_tables_data.clear();
2020-06-02 16:27:05 +00:00
external_tables_data.reserve(count);
StreamLocalLimits limits;
const auto & settings = context->getSettingsRef();
limits.mode = LimitsMode::LIMITS_TOTAL;
limits.speed_limits.max_execution_time = settings.max_execution_time;
limits.timeout_overflow_mode = settings.timeout_overflow_mode;
2020-06-02 16:27:05 +00:00
for (size_t i = 0; i < count; ++i)
{
ExternalTablesData res;
for (const auto & table : external_tables)
{
StoragePtr cur = table.second;
/// Send only temporary tables with StorageMemory
if (!std::dynamic_pointer_cast<StorageMemory>(cur))
continue;
2020-06-02 16:27:05 +00:00
auto data = std::make_unique<ExternalTableData>();
data->table_name = table.first;
data->creating_pipe_callback = [cur, limits, context = this->context]()
{
SelectQueryInfo query_info;
auto metadata_snapshot = cur->getInMemoryMetadataPtr();
auto storage_snapshot = cur->getStorageSnapshot(metadata_snapshot, context);
QueryProcessingStage::Enum read_from_table_stage = cur->getQueryProcessingStage(
context, QueryProcessingStage::Complete, storage_snapshot, query_info);
2022-05-23 19:47:32 +00:00
QueryPlan plan;
cur->read(
plan,
metadata_snapshot->getColumns().getNamesOfPhysical(),
storage_snapshot, query_info, context,
read_from_table_stage, DEFAULT_BLOCK_SIZE, 1);
2022-05-23 19:47:32 +00:00
auto builder = plan.buildQueryPipeline(
QueryPlanOptimizationSettings::fromContext(context),
BuildQueryPipelineSettings::fromContext(context));
2020-06-02 16:27:05 +00:00
2022-05-23 19:47:32 +00:00
builder->resize(1);
builder->addTransform(std::make_shared<LimitsCheckingTransform>(builder->getHeader(), limits));
2022-05-23 19:47:32 +00:00
return builder;
};
data->pipe = data->creating_pipe_callback();
2020-06-02 16:27:05 +00:00
res.emplace_back(std::move(data));
}
external_tables_data.push_back(std::move(res));
}
}
2021-01-19 19:21:06 +00:00
connections->sendExternalTablesData(external_tables_data);
2020-06-02 16:27:05 +00:00
}
2020-12-17 10:07:28 +00:00
void RemoteQueryExecutor::tryCancel(const char * reason, std::unique_ptr<ReadContext> * read_context)
2020-06-02 16:27:05 +00:00
{
Fix possible "No more packets are available" for distributed queries CI founds the following case: <details> 2022.05.25 22:36:06.778808 [ 3037 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Fatal> : Logical error: 'No more packets are available.'. 2022.05.25 22:42:24.960075 [ 17397 ] {} <Fatal> BaseDaemon: ######################################## 2022.05.25 22:42:24.971173 [ 17397 ] {} <Fatal> BaseDaemon: (version 22.6.1.1, build id: 9A1F9489854CED36) (from thread 3037) (query_id: 77743723-1fcd-4b3d-babc-d0615e3ff40e) (query: SELECT * FROM 2022.05.25 22:42:25.046871 [ 17397 ] {} <Fatal> BaseDaemon: 5. ./build_docker/../src/Common/Exception.cpp:47: DB::abortOnFailedAssertion() 2022.05.25 22:42:25.181449 [ 17397 ] {} <Fatal> BaseDaemon: 6. ./build_docker/../src/Common/Exception.cpp:70: DB::Exception::Exception() 2022.05.25 22:42:25.367710 [ 17397 ] {} <Fatal> BaseDaemon: 7. ./build_docker/../src/Client/MultiplexedConnections.cpp:0: DB::MultiplexedConnections::receivePacketUnlocked() 2022.05.25 22:42:25.414201 [ 17397 ] {} <Fatal> BaseDaemon: 8. ./build_docker/../src/Client/MultiplexedConnections.cpp:0: DB::MultiplexedConnections::receivePacket() 2022.05.25 22:42:25.493066 [ 17397 ] {} <Fatal> BaseDaemon: 9. ./build_docker/../src/QueryPipeline/RemoteQueryExecutor.cpp:279: DB::RemoteQueryExecutor::read() 2022.05.25 22:42:25.612679 [ 17397 ] {} <Fatal> BaseDaemon: 10. ./build_docker/../src/Processors/Sources/RemoteSource.cpp:0: DB::RemoteSource::tryGenerate() Here are additional logs for this query: $ pigz -cd clickhouse-server.stress.log.gz | fgrep -a 77743723-1fcd-4b3d-babc-d0615e3ff40e | fgrep -e Connection -e Distributed -e Fatal 2022.05.25 22:36:04.698671 [ 6613 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Trace> Connection (127.0.0.2:9000): Connecting. Database: (not specified). User: default 2022.05.25 22:36:04.722568 [ 3419 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Trace> Connection (127.0.0.2:9000): Connecting. Database: (not specified). User: default 2022.05.25 22:36:05.014432 [ 6613 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Trace> Connection (127.0.0.2:9000): Connected to ClickHouse server version 22.6.1. 2022.05.25 22:36:05.091397 [ 6613 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Debug> Connection (127.0.0.2:9000): Sent data for 2 scalars, total 2 rows in 0.000125814 sec., 15602 rows/sec., 68.00 B (517.81 KiB/sec.), compressed 0.4594594594594595 times to 148.00 B (1.10 MiB/sec.) 2022.05.25 22:36:05.301301 [ 3419 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Trace> Connection (127.0.0.2:9000): Connected to ClickHouse server version 22.6.1. 2022.05.25 22:36:05.343140 [ 3419 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Debug> Connection (127.0.0.2:9000): Sent data for 2 scalars, total 2 rows in 0.000116304 sec., 16889 rows/sec., 68.00 B (559.80 KiB/sec.), compressed 0.4594594594594595 times to 148.00 B (1.19 MiB/sec.) 2022.05.25 22:36:06.682535 [ 6613 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Trace> StorageDistributed (remote): (127.0.0.2:9000) Cancelling query because enough data has been read 2022.05.25 22:36:06.778808 [ 3037 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Fatal> : Logical error: 'No more packets are available.'. 2022.05.25 22:36:06.789505 [ 3419 ] {77743723-1fcd-4b3d-babc-d0615e3ff40e} <Trace> StorageDistributed (remote): (127.0.0.2:9000) Cancelling query because enough data has been read 2022.05.25 22:42:24.971173 [ 17397 ] {} <Fatal> BaseDaemon: (version 22.6.1.1, build id: 9A1F9489854CED36) (from thread 3037) (query_id: 77743723-1fcd-4b3d-babc-d0615e3ff40e) (query: SELECT * FROM </details> So between cancelling different sources the LOGICAL_ERROR occured, I believe that this is because of the race: T1: T2: RemoteQueryExecutor::read() checks was_cancelled RemoteQueryExecutor::tryCancel() connections->cancel() calls connections->receivePacket() Note, for this problem async_socket_for_remote/use_hedged_requests should be disabled, and original settings was: - --max_parallel_replicas=3 - --use_hedged_requests=false - --allow_experimental_parallel_reading_from_replicas=3 CI: https://s3.amazonaws.com/clickhouse-test-reports/37469/41cb029ed23e77f3a108e07e6b1b1bcb03dc7fcf/stress_test__undefined__actions_/fatal_messages.txt Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com>
2022-06-03 11:49:22 +00:00
/// Flag was_cancelled is atomic because it is checked in read(),
/// in case of packet had been read by fiber (async_socket_for_remote).
2021-07-14 13:17:30 +00:00
std::lock_guard guard(was_cancelled_mutex);
2020-06-02 16:27:05 +00:00
2021-07-14 13:17:30 +00:00
if (was_cancelled)
return;
2020-06-02 16:27:05 +00:00
2021-07-14 13:17:30 +00:00
was_cancelled = true;
2020-12-17 10:07:28 +00:00
2021-07-14 13:17:30 +00:00
if (read_context && *read_context)
{
/// The timer should be set for query cancellation to avoid query cancellation hung.
///
/// Since in case the remote server will abnormally terminated, neither
/// FIN nor RST packet will be sent, and the initiator will not know that
/// the connection died (unless tcp_keep_alive_timeout > 0).
///
/// Also note that it is possible to get this situation even when
/// enough data already had been read.
(*read_context)->setTimer();
2021-07-14 13:17:30 +00:00
(*read_context)->cancel();
}
2020-12-17 10:07:28 +00:00
2021-07-14 13:17:30 +00:00
connections->sendCancel();
2020-06-02 16:27:05 +00:00
if (log)
2021-01-19 19:21:06 +00:00
LOG_TRACE(log, "({}) {}", connections->dumpAddresses(), reason);
2020-06-02 16:27:05 +00:00
}
bool RemoteQueryExecutor::isQueryPending() const
{
return sent_query && !finished;
}
bool RemoteQueryExecutor::hasThrownException() const
{
return got_exception_from_replica || got_unknown_packet_from_replica;
}
}