自省函数
您可以使用本章介绍的函数对 ELF 和 DWARF 进行自省，以进行查询性能剖析。
注意
这些函数执行较慢，并且可能带来安全风险。
要确保自省函数正常工作：
-
安装
clickhouse-common-static-dbg软件包。 -
将 allow_introspection_functions 设置为 1。
出于安全原因，自省函数默认处于禁用状态。
ClickHouse 会将分析器生成的报告保存到 trace_log 系统表中。请确保该表和分析器已正确配置。
addressToLine
引入版本：v20.1
将 ClickHouse 服务器进程中的虚拟内存地址转换为 ClickHouse 源代码中的文件名和行号。
注意
此函数较慢，并且可能带来安全方面的风险。
要启用此自省函数：
- 安装
clickhouse-common-static-dbg软件包。 - 将设置
allow_introspection_functions设为1。
语法
addressToLine(address_of_binary_instruction)
参数
address_of_binary_instruction— 正在运行的进程中某条指令的地址。UInt64
返回值
返回以冒号分隔的源代码文件名和行号，例如 /build/obj-x86_64-linux-gnu/../src/Common/ThreadPool.cpp:199。如果找不到调试信息，则返回二进制文件名；如果地址无效，则返回空字符串。String
示例
从 trace_log 系统表中选取第一条字符串
SET allow_introspection_functions=1;
SELECT * FROM system.trace_log LIMIT 1 \G;
-- `trace` 字段包含采样时刻的堆栈跟踪信息。
行 1:
──────
event_date: 2019-11-19
event_time: 2019-11-19 18:57:23
revision: 54429
timer_type: Real
thread_number: 48
query_id: 421b6855-1858-45a5-8f37-f383409d6d72
trace: [140658411141617,94784174532828,94784076370703,94784076372094,94784076361020,94784175007680,140658411116251,140658403895439]
获取单个地址对应的源代码文件名和行号
SET allow_introspection_functions=1;
SELECT addressToLine(94784076370703) \G;
Row 1:
──────
addressToLine(94784076370703): /build/obj-x86_64-linux-gnu/../src/Common/ThreadPool.cpp:199
将该函数应用于完整堆栈跟踪
-- 此示例中的 arrayMap 函数通过 addressToLine 函数处理 trace 数组中的每个元素。
-- 处理结果显示在输出的 trace_source_code_lines 列中。
SELECT
arrayStringConcat(arrayMap(x -> addressToLine(x), trace), '\n') AS trace_source_code_lines
FROM system.trace_log
LIMIT 1
\G
行 1:
──────
trace_source_code_lines: /lib/x86_64-linux-gnu/libpthread-2.27.so
/usr/lib/debug/usr/bin/clickhouse
/build/obj-x86_64-linux-gnu/../src/Common/ThreadPool.cpp:199
/build/obj-x86_64-linux-gnu/../src/Common/ThreadPool.h:155
/usr/include/c++/9/bits/atomic_base.h:551
/usr/lib/debug/usr/bin/clickhouse
/lib/x86_64-linux-gnu/libpthread-2.27.so
/build/glibc-OTsEL5/glibc-2.27/misc/../sysdeps/unix/sysv/linux/x86_64/clone.S:97
addressToLineWithInlines
引入于：v22.2
与 addressToLine 类似，但返回包含所有内联函数的数组（Array）。
因此，它比 addressToLine 更慢。
要启用此自省函数：
- 安装
clickhouse-common-static-dbg软件包。 - 将设置项
allow_introspection_functions设为1。
语法
addressToLineWithInlines(address_of_binary_instruction)
参数
address_of_binary_instruction— 正在运行的进程中某条指令的地址。UInt64
返回值
返回一个数组，第一个元素为源代码文件名和行号，中间以冒号分隔。第二、第三等后续元素用于列出内联函数的源代码文件名、行号和函数名。如果找不到调试信息，则返回仅包含一个元素（等于二进制可执行文件名称）的数组；如果地址无效，则返回空数组。Array(String)
示例
对某个地址应用该函数
SET allow_introspection_functions=1;
SELECT addressToLineWithInlines(531055181::UInt64);
┌─addressToLineWithInlines(CAST('531055181', 'UInt64'))────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│ ['./src/Functions/addressToLineWithInlines.cpp:98','./build_normal_debug/./src/Functions/addressToLineWithInlines.cpp:176:DB::(anonymous namespace)::FunctionAddressToLineWithInlines::implCached(unsigned long) const'] │
└──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┘
将该函数应用于整个堆栈跟踪
SET allow_introspection_functions=1;
-- arrayJoin 函数将数组拆分为多行
SELECT
ta, addressToLineWithInlines(arrayJoin(trace) AS ta)
FROM system.trace_log
WHERE
query_id = '5e173544-2020-45de-b645-5deebe2aae54';
┌────────ta─┬─addressToLineWithInlines(arrayJoin(trace))───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│ 365497529 │ ['./build_normal_debug/./contrib/libcxx/include/string_view:252'] │
│ 365593602 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:191'] │
│ 365593866 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:0'] │
│ 365592528 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:0'] │
│ 365591003 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:477'] │
│ 365590479 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:442'] │
│ 365590600 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:457'] │
│ 365598941 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:0'] │
│ 365607098 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:0'] │
│ 365590571 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:451'] │
│ 365598941 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:0'] │
│ 365607098 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:0'] │
│ 365590571 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:451'] │
│ 365598941 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:0'] │
│ 365607098 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:0'] │
│ 365590571 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:451'] │
│ 365598941 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:0'] │
│ 365597289 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:807'] │
│ 365599840 │ ['./build_normal_debug/./src/Common/Dwarf.cpp:1118'] │
│ 531058145 │ ['./build_normal_debug/./src/Functions/addressToLineWithInlines.cpp:152'] │
│ 531055181 │ ['./src/Functions/addressToLineWithInlines.cpp:98','./build_normal_debug/./src/Functions/addressToLineWithInlines.cpp:176:DB::(anonymous namespace)::FunctionAddressToLineWithInlines::implCached(unsigned long) const'] │
│ 422333613 │ ['./build_normal_debug/./src/Functions/IFunctionAdaptors.h:21'] │
│ 586866022 │ ['./build_normal_debug/./src/Functions/IFunction.cpp:216'] │
│ 586869053 │ ['./build_normal_debug/./src/Functions/IFunction.cpp:264'] │
│ 586873237 │ ['./build_normal_debug/./src/Functions/IFunction.cpp:334'] │
│ 597901620 │ ['./build_normal_debug/./src/Interpreters/ExpressionActions.cpp:601'] │
│ 597898534 │ ['./build_normal_debug/./src/Interpreters/ExpressionActions.cpp:718'] │
│ 630442912 │ ['./build_normal_debug/./src/Processors/Transforms/ExpressionTransform.cpp:23'] │
│ 546354050 │ ['./build_normal_debug/./src/Processors/ISimpleTransform.h:38'] │
│ 626026993 │ ['./build_normal_debug/./src/Processors/ISimpleTransform.cpp:89'] │
│ 626294022 │ ['./build_normal_debug/./src/Processors/Executors/ExecutionThreadContext.cpp:45'] │
│ 626293730 │ ['./build_normal_debug/./src/Processors/Executors/ExecutionThreadContext.cpp:63'] │
│ 626169525 │ ['./build_normal_debug/./src/Processors/Executors/PipelineExecutor.cpp:213'] │
│ 626170308 │ ['./build_normal_debug/./src/Processors/Executors/PipelineExecutor.cpp:178'] │
│ 626166348 │ ['./build_normal_debug/./src/Processors/Executors/PipelineExecutor.cpp:329'] │
│ 626163461 │ ['./build_normal_debug/./src/Processors/Executors/PipelineExecutor.cpp:84'] │
│ 626323536 │ ['./build_normal_debug/./src/Processors/Executors/PullingAsyncPipelineExecutor.cpp:85'] │
│ 626323277 │ ['./build_normal_debug/./src/Processors/Executors/PullingAsyncPipelineExecutor.cpp:112'] │
│ 626323133 │ ['./build_normal_debug/./contrib/libcxx/include/type_traits:3682'] │
│ 626323041 │ ['./build_normal_debug/./contrib/libcxx/include/tuple:1415'] │
└───────────┴──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┘
addressToSymbol
引入版本：v20.1
将 ClickHouse 服务器进程中的虚拟内存地址转换为 ClickHouse 对象文件中的符号。
语法
addressToSymbol(address_of_binary_instruction)
参数
address_of_binary_instruction— 正在运行的进程中指令的地址。UInt64
返回值
返回 ClickHouse 对象文件中的符号，如果地址无效则返回空字符串。String
示例
从 trace_log 系统表中选取第一条记录
SET allow_introspection_functions=1;
SELECT * FROM system.trace_log LIMIT 1 \G;
-- `trace` 字段包含采样时刻的堆栈跟踪。
行 1:
──────
event_date: 2019-11-20
event_time: 2019-11-20 16:57:59
revision: 54429
timer_type: Real
thread_number: 48
query_id: 724028bf-f550-45aa-910d-2af6212b94ac
trace: [94138803686098,94138815010911,94138815096522,94138815101224,94138815102091,94138814222988,94138806823642,94138814457211,94138806823642,94138814457211,94138806823642,94138806795179,94138806796144,94138753770094,94138753771646,94138753760572,94138852407232,140399185266395,140399178045583]
获取单个地址的符号信息
SET allow_introspection_functions=1;
SELECT addressToSymbol(94138803686098) \G;
Row 1:
──────
addressToSymbol(94138803686098): _ZNK2DB24IAggregateFunctionHelperINS_20AggregateFunctionSumImmNS_24AggregateFunctionSumDataImEEEEE19addBatchSinglePlaceEmPcPPKNS_7IColumnEPNS_5ArenaE
将该函数应用于整个堆栈跟踪
SET allow_introspection_functions=1;
-- arrayMap 函数允许通过 addressToSymbol 函数处理 trace 数组中的每个元素。
-- 处理结果显示在输出的 trace_symbols 列中。
SELECT
arrayStringConcat(arrayMap(x -> addressToSymbol(x), trace), '\n') AS trace_symbols
FROM system.trace_log
LIMIT 1
\G
Row 1:
──────
trace_symbols: _ZNK2DB24IAggregateFunctionHelperINS_20AggregateFunctionSumImmNS_24AggregateFunctionSumDataImEEEEE19addBatchSinglePlaceEmPcPPKNS_7IColumnEPNS_5ArenaE
_ZNK2DB10Aggregator21executeWithoutKeyImplERPcmPNS0_28AggregateFunctionInstructionEPNS_5ArenaE
_ZN2DB10Aggregator14executeOnBlockESt6vectorIN3COWINS_7IColumnEE13immutable_ptrIS3_EESaIS6_EEmRNS_22AggregatedDataVariantsERS1_IPKS3_SaISC_EERS1_ISE_SaISE_EERb
_ZN2DB10Aggregator14executeOnBlockERKNS_5BlockERNS_22AggregatedDataVariantsERSt6vectorIPKNS_7IColumnESaIS9_EERS6_ISB_SaISB_EERb
_ZN2DB10Aggregator7executeERKSt10shared_ptrINS_17IBlockInputStreamEERNS_22AggregatedDataVariantsE
_ZN2DB27AggregatingBlockInputStream8readImplEv
_ZN2DB17IBlockInputStream4readEv
_ZN2DB26ExpressionBlockInputStream8readImplEv
_ZN2DB17IBlockInputStream4readEv
_ZN2DB26ExpressionBlockInputStream8readImplEv
_ZN2DB17IBlockInputStream4readEv
_ZN2DB28AsynchronousBlockInputStream9calculateEv
_ZNSt17_Function_handlerIFvvEZN2DB28AsynchronousBlockInputStream4nextEvEUlvE_E9_M_invokeERKSt9_Any_data
_ZN14ThreadPoolImplI20ThreadFromGlobalPoolE6workerESt14_List_iteratorIS0_E
_ZZN20ThreadFromGlobalPoolC4IZN14ThreadPoolImplIS_E12scheduleImplIvEET_St8functionIFvvEEiSt8optionalImEEUlvE1_JEEEOS4_DpOT0_ENKUlvE_clEv
_ZN14ThreadPoolImplISt6threadE6workerESt14_List_iteratorIS0_E
execute_native_thread_routine
start_thread
clone
demangle
引入自：v20.1 版本
将符号转换为 C++ 函数名。
该符号通常由函数 addressToSymbol 返回。
语法
demangle(symbol)
参数
symbol— 来自对象文件的符号。String
返回值
返回 C++ 函数的名称，如果符号无效则返回空字符串。String
示例
从 trace_log 系统表中选取第一个字符串
SELECT * FROM system.trace_log LIMIT 1 \G;
-- `trace` 字段包含采样时刻的堆栈跟踪。
行 1:
──────
event_date: 2019-11-20
event_time: 2019-11-20 16:57:59
revision: 54429
timer_type: Real
thread_number: 48
query_id: 724028bf-f550-45aa-910d-2af6212b94ac
trace: [94138803686098,94138815010911,94138815096522,94138815101224,94138815102091,94138814222988,94138806823642,94138814457211,94138806823642,94138814457211,94138806823642,94138806795179,94138806796144,94138753770094,94138753771646,94138753760572,94138852407232,140399185266395,140399178045583]
获取单个地址对应的函数名
SET allow_introspection_functions=1;
SELECT demangle(addressToSymbol(94138803686098)) \G;
Row 1:
──────
demangle(addressToSymbol(94138803686098)): DB::IAggregateFunctionHelper<DB::AggregateFunctionSum<unsigned long, unsigned long, DB::AggregateFunctionSumData<unsigned long> > >::addBatchSinglePlace(unsigned long, char*, DB::IColumn const**, DB::Arena*) const
将函数应用于整个堆栈跟踪
SET allow_introspection_functions=1;
-- arrayMap 函数允许使用 demangle 函数处理 trace 数组中的每个元素。
-- 处理结果显示在输出的 trace_functions 列中。
SELECT
arrayStringConcat(arrayMap(x -> demangle(addressToSymbol(x)), trace), '\n') AS trace_functions
FROM system.trace_log
LIMIT 1
\G
Row 1:
──────
trace_functions: DB::IAggregateFunctionHelper<DB::AggregateFunctionSum<unsigned long, unsigned long, DB::AggregateFunctionSumData<unsigned long> > >::addBatchSinglePlace(unsigned long, char*, DB::IColumn const**, DB::Arena*) const
DB::Aggregator::executeWithoutKeyImpl(char*&, unsigned long, DB::Aggregator::AggregateFunctionInstruction*, DB::Arena*) const
DB::Aggregator::executeOnBlock(std::vector<COW<DB::IColumn>::immutable_ptr<DB::IColumn>, std::allocator<COW<DB::IColumn>::immutable_ptr<DB::IColumn> > >, unsigned long, DB::AggregatedDataVariants&, std::vector<DB::IColumn const*, std::allocator<DB::IColumn const*> >&, std::vector<std::vector<DB::IColumn const*, std::allocator<DB::IColumn const*> >, std::allocator<std::vector<DB::IColumn const*, std::allocator<DB::IColumn const*> > > >&, bool&)
DB::Aggregator::executeOnBlock(DB::Block const&, DB::AggregatedDataVariants&, std::vector<DB::IColumn const*, std::allocator<DB::IColumn const*> >&, std::vector<std::vector<DB::IColumn const*, std::allocator<DB::IColumn const*> >, std::allocator<std::vector<DB::IColumn const*, std::allocator<DB::IColumn const*> > > >&, bool&)
DB::Aggregator::execute(std::shared_ptr<DB::IBlockInputStream> const&, DB::AggregatedDataVariants&)
DB::AggregatingBlockInputStream::readImpl()
DB::IBlockInputStream::read()
DB::ExpressionBlockInputStream::readImpl()
DB::IBlockInputStream::read()
DB::ExpressionBlockInputStream::readImpl()
DB::IBlockInputStream::read()
DB::AsynchronousBlockInputStream::calculate()
std::_Function_handler<void (), DB::AsynchronousBlockInputStream::next()::{lambda()#1}>::_M_invoke(std::_Any_data const&)
ThreadPoolImpl<ThreadFromGlobalPool>::worker(std::_List_iterator<ThreadFromGlobalPool>)
ThreadFromGlobalPool::ThreadFromGlobalPool<ThreadPoolImpl<ThreadFromGlobalPool>::scheduleImpl<void>(std::function<void ()>, int, std::optional<unsigned long>)::{lambda()#3}>(ThreadPoolImpl<ThreadFromGlobalPool>::scheduleImpl<void>(std::function<void ()>, int, std::optional<unsigned long>)::{lambda()#3}&&)::{lambda()#1}::operator()() const
ThreadPoolImpl<std::thread>::worker(std::_List_iterator<std::thread>)
execute_native_thread_routine
start_thread
clone
isMergeTreePartCoveredBy
在 v25.6 中引入
检查第一个参数所表示的数据分片是否被第二个参数所表示的数据分片覆盖的 FUNCTION。
语法
isMergeTreePartCoveredBy(nested_part, covering_part)
参数
返回值
如果满足覆盖关系则返回 1，否则返回 0。UInt8
示例
基本示例
WITH 'all_12_25_7_4' AS lhs, 'all_7_100_10_20' AS rhs
SELECT isMergeTreePartCoveredBy(rhs, lhs), isMergeTreePartCoveredBy(lhs, rhs);
┌─isMergeTreePartCoveredBy(rhs, lhs)─┬─isMergeTreePartCoveredBy(lhs, rhs)─┐
│ 0 │ 1 │
└────────────────────────────────────┴────────────────────────────────────┘
logTrace
自 v20.12 引入
为每个 Block 在服务器日志中写入一条跟踪日志记录。
语法
logTrace(message)
参数
message— 写入服务器日志的消息。const String
返回值
始终返回 0。UInt8
示例
基本示例
SELECT logTrace('logTrace message');
┌─logTrace('logTrace message')─┐
│ 0 │
└──────────────────────────────┘
mergeTreePartInfo
引入版本：v25.6
用于从 MergeTree 数据部分名称中提取有用信息的函数。
语法
mergeTreePartInfo(part_name)
参数
part_name— 要解包的 part 名称。String
返回值
返回一个 Tuple，其中包含以下子列：partition_id、min_block、max_block、level、mutation。Tuple
示例
基本示例
WITH mergeTreePartInfo('all_12_25_7_4') AS info
SELECT info.partition_id, info.min_block, info.max_block, info.level, info.mutation;
┌─info.partition_id─┬─info.min_block─┬─info.max_block─┬─info.level─┬─info.mutation─┐
│ all │ 12 │ 25 │ 7 │ 4 │
└───────────────────┴────────────────┴────────────────┴────────────┴───────────────┘
tid
引入版本：v20.12
返回处理当前 Block 的线程 ID。
语法
tid()
参数
- 无。
返回值
返回当前线程 ID。UInt64
示例
使用示例
SELECT tid();
┌─tid()─┐
│ 3878 │
└───────┘
