Starrocks中的 Query Profile以及explain analyze及trace命令中的区别

背景

本文基于Starrocks 3.5.5

现有公司因为业务的不同，可能会更加关系单个SQL 的RT，因为如果一个SQL的RT比较大的话，影响的就是这个业务，从而影响收入，所以对于这方面我们就比较关心，

而最近在基于Starrocks做计算存储引擎的时候，遇到了一些问题。时间上超过了2秒，因此需要分析一下对应SQL的指标。

最直接的就是开启 Query Profile,比如说做一下配置:

SET enable_profile = true;
SET global big_query_profile_threshold = '1s';
SET runtime_profile_report_interval = 30;

这个在对应的UI界面（http://<fe_ip>:<fe_http_port>）就能看到Profile

当然还有ANALYZE PROFILE,explain analyze和trace命令,这里就来分析一下三者的差别， 其中以以下SQL为例：

SELECT `SCHEMA_NAME` FROM `INFORMATION_SCHEMA`.`SCHEMATA` WHERE SCHEMA_NAME = 'fin_config';

结论

1. 开启Query Profile会得得到更多的明细信息，以及每个阶段所用的耗时,适合对该SQL的每个算子的各个指标全方位的分析
   此种方法会有BE端的信息更新过来，但是这个信息只存在内存，FE重启之后就不复存在。
2. EXPLAIN ANALYZE 展示的是大概的查询执行计划执行信息，比如说 Summary和fragement
3. ANALYZE PROFILE 展示的和explain analyze一样，但是指标比 explain analyze 更加丰富
4. trace 展示的是某个局部指标，比如说 某个规则的耗时

对于每个方法所对应指标信息如下见下面的 其他

分析

开启Query Profile

通过访问http://<fe_ip>:<fe_http_port>/query_profile?query_id可以获取到对应的profile，

对应到Fe端是QueryProfileAction类接受请求：

QueryProfileAction 里有对应的 "/query_profile" Get请求

对应的Get方法response为executeGet方法：

String queryProfileStr = ProfileManager.getInstance().getProfile(queryId);
        if (queryProfileStr != null) {
            appendCopyButton(response.getContent());
            appendQueryProfile(response.getContent(), queryProfileStr);
            getPageFooter(response.getContent());
            writeResponse(request, response);

这里ProfileManager.getInstance().getProfile(queryId)会获取对应的 Profile.

runtimeProfile数据流

FrontendServiceImpl.reportExecStatus
    ||
    \/
DefaultCoordinator.updateFragmentExecStatus(params)
    ||
    \/
QueryRuntimeProfile.updateProfile(execState, params) // 这里会 runtime_profile_report_interval 判断
    ||
    \/
saveRunningProfile
    ||
    \/
ProfileManager.pushProfile(profilingPlan, profile);

这里会把 BE端 Fragment 实力的 运行指标都给返回给 FE端.

EXPLAIN ANALYZE

具体语法参考Simulate a query for Profile Analysis Using EXPLAIN ANALYZE

直接转到StarRocks.g4

queryStatement
    : (explainDesc | optimizerTrace) ? queryRelation outfile?;
...
explainDesc
    : (DESC | DESCRIBE | EXPLAIN) (LOGICAL | ANALYZE | VERBOSE | COSTS | SCHEDULER)?
    ;
...
optimizerTrace
    : TRACE (ALL | LOGS | TIMES | VALUES | REASON) identifier?
    ;

通过AstBuilder.visitQueryStatement 解析完后:

@Override
public ParseNode visitQueryStatement(StarRocksParser.QueryStatementContext context) {
    QueryRelation queryRelation = (QueryRelation) visit(context.queryRelation());
    QueryStatement queryStatement = new QueryStatement(queryRelation);
    if (context.outfile() != null) {
        queryStatement.setOutFileClause((OutFileClause) visit(context.outfile()));
    }

    if (context.explainDesc() != null) {
        queryStatement.setIsExplain(true, getExplainType(context.explainDesc()));
    }

    if (context.optimizerTrace() != null) {
        String module = "base";
        if (context.optimizerTrace().identifier() != null) {
            module = ((Identifier) visit(context.optimizerTrace().identifier())).getValue();
        }
        queryStatement.setIsTrace(getTraceMode(context.optimizerTrace()), module);
    }

    return queryStatement;
}

可以看到queryStatement.setIsExplain(true, getExplainType(context.explainDesc()))这个方法：

  public void setIsExplain(boolean isExplain, ExplainLevel explainLevel) {
        this.isExplain = isExplain;
        this.explainLevel = explainLevel;
    }

会把isExplain设置为true,

之后数据流会转到StmtExecutor.execute方法,最终会调用handleQueryStmt方法：

private void handleQueryStmt(ExecPlan execPlan) throws Exception {
        // Every time set no send flag and clean all data in buffer
        context.getMysqlChannel().reset();

        boolean isExplainAnalyze = parsedStmt.isExplain()
                && StatementBase.ExplainLevel.ANALYZE.equals(parsedStmt.getExplainLevel());
        boolean isSchedulerExplain = parsedStmt.isExplain()
                && StatementBase.ExplainLevel.SCHEDULER.equals(parsedStmt.getExplainLevel());
        boolean isOutfileQuery = (parsedStmt instanceof QueryStatement) && ((QueryStatement) parsedStmt).hasOutFileClause();
        if (isOutfileQuery) {
            Map<TableName, Table> tables = AnalyzerUtils.collectAllTable(parsedStmt);
            boolean hasTemporaryTable = tables.values().stream().anyMatch(t -> t.isTemporaryTable());
            if (hasTemporaryTable) {
                throw new SemanticException("temporary table doesn't support select outfile statement");
            }
        }
        boolean executeInFe = !isExplainAnalyze && !isSchedulerExplain && !isOutfileQuery
                && canExecuteInFe(context, execPlan.getPhysicalPlan());

        if (isExplainAnalyze) {
            context.getSessionVariable().setEnableProfile(true);
            context.getSessionVariable().setEnableAsyncProfile(false);
            context.getSessionVariable().setPipelineProfileLevel(1);
        } else if (isSchedulerExplain) {
            // Do nothing.
        } else if (parsedStmt.isExplain()) {
            String explainString = buildExplainString(execPlan, ResourceGroupClassifier.QueryType.SELECT,
                    parsedStmt.getExplainLevel());
            if (executeInFe) {
                explainString = "EXECUTE IN FE\n" + explainString;
            }
            handleExplainStmt(explainString);
            return;
        }
        ... 

        StatementBase queryStmt = parsedStmt;
        List<PlanFragment> fragments = execPlan.getFragments();
        List<ScanNode> scanNodes = execPlan.getScanNodes();
        TDescriptorTable descTable = execPlan.getDescTbl().toThrift();
        List<String> colNames = execPlan.getColNames();
        List<Expr> outputExprs = execPlan.getOutputExprs();

        if (executeInFe) {
            coord = new FeExecuteCoordinator(context, execPlan);
        } else {
            coord = getCoordinatorFactory().createQueryScheduler(context, fragments, scanNodes, descTable);
        }

        QeProcessorImpl.INSTANCE.registerQuery(context.getExecutionId(),
                new QeProcessorImpl.QueryInfo(context, originStmt.originStmt, coord));

        if (isSchedulerExplain) {
            coord.startSchedulingWithoutDeploy();
            handleExplainStmt(coord.getSchedulerExplain());
            return;
        }

        coord.exec();
        coord.setTopProfileSupplier(this::buildTopLevelProfile);
        coord.setExecPlan(execPlan);

对于 explain analyze 会进行如下配置:

 context.getSessionVariable().setEnableProfile(true);
 context.getSessionVariable().setEnableAsyncProfile(false);
 context.getSessionVariable().setPipelineProfileLevel(1);

且会走到 coord.exec();这里会有调度的部分，如以下：

 try (Timer timer = Tracers.watchScope(Tracers.Module.SCHEDULER, "Prepare")) {
            prepareExec();
        }

 try (Timer timer = Tracers.watchScope(Tracers.Module.SCHEDULER, "Deploy")) {
     deliverExecFragments(needDeploy);
 }

而在handleQueryStmt下后有finally的处理：

 else if (context.isProfileEnabled()) {
    isAsync = tryProcessProfileAsync(execPlan, i);
    if (parsedStmt.isExplain() &&
            StatementBase.ExplainLevel.ANALYZE.equals(parsedStmt.getExplainLevel())) {
        if (coord != null && coord.isShortCircuit()) {
            throw new UserException(
                    "short circuit point query doesn't suppot explain analyze stmt, " +
                            "you can set it off by using  set enable_short_circuit=false");
        }
        handleExplainStmt(ExplainAnalyzer.analyze(
                ProfilingExecPlan.buildFrom(execPlan), profile, null));
    }
 }

ExplainAnalyzer.analyze这里就是返回的Explain String

ANALYZE PROFILE

直接跳转到Starrocks.g4:

analyzeProfileStatement
    : ANALYZE PROFILE FROM string
    | ANALYZE PROFILE FROM string ',' INTEGER_VALUE (',' INTEGER_VALUE)*
    ;

之后到AstBuilder的如下方法:

@Override
public ParseNode visitAnalyzeProfileStatement(StarRocksParser.AnalyzeProfileStatementContext context) {
    StringLiteral stringLiteral = (StringLiteral) visit(context.string());
    List<Integer> planNodeIds = Lists.newArrayList();
    if (context.INTEGER_VALUE() != null) {
        planNodeIds = context.INTEGER_VALUE().stream()
                .map(ParseTree::getText)
                .map(Integer::parseInt)
                .collect(toList());
    }
    return new AnalyzeProfileStmt(stringLiteral.getStringValue(), planNodeIds, createPos(context));
}

之后再到StmtExecutor.handleAnalyzeProfileStmt方法:

private void handleAnalyzeProfileStmt() throws IOException, UserException {
        AnalyzeProfileStmt analyzeProfileStmt = (AnalyzeProfileStmt) parsedStmt;
        String queryId = analyzeProfileStmt.getQueryId();
        List<Integer> planNodeIds = analyzeProfileStmt.getPlanNodeIds();
        ProfileManager.ProfileElement profileElement = ProfileManager.getInstance().getProfileElement(queryId);
        Preconditions.checkNotNull(profileElement, "query not exists");
        // For short circuit query, 'ProfileElement#plan' is null
        if (profileElement.plan == null) {
            throw new UserException(
                    "short circuit point query doesn't suppot analyze profile stmt, " +
                            "you can set it off by using  set enable_short_circuit=false");
        }
        handleExplainStmt(ExplainAnalyzer.analyze(profileElement.plan,
                RuntimeProfileParser.parseFrom(CompressionUtils.gzipDecompressString(profileElement.profileContent)),
                planNodeIds));
    }

这里通过ProfileManager.getInstance().getProfileElement(queryId)获取到对应的profile,

之后再通过ExplainAnalyzer.analyze获取对应的explain string.

TRACE

具体语法参考query_trace_profile

和 EXPLAIN ANALYZE的逻辑一样，通过AstBuilder.visitQueryStatement 解析完后,走的是queryStatement.setIsTrace(getTraceMode(context.optimizerTrace()), module);这个逻辑：

 public void setIsTrace(Tracers.Mode mode, String module) {
        this.isExplain = true;
        this.traceMode = mode;
        this.traceModule = module;
    }

可以看到这里的isExplain为true.

而handleQueryStmt方法为:

} else if (parsedStmt.isExplain()) {
    String explainString = buildExplainString(execPlan, ResourceGroupClassifier.QueryType.SELECT,
            parsedStmt.getExplainLevel());
    if (executeInFe) {
        explainString = "EXECUTE IN FE\n" + explainString;
    }
    handleExplainStmt(explainString);
    return;
}

这里的buildExplainString方法，会根据trace的mode来进行explain String的构建。

其他

1. 开启Query Profile 指标

   | Query:
   Summary:
   - Query ID: 135d8852-62fd-11f0-b356-00163e164034
   - Start Time: 2025-07-17 18:59:13
   - End Time: 2025-07-17 18:59:14
   - Total: 268ms
   - Query Type: Query
   - Query State: Finished
   - StarRocks Version: 3.3.5-6d81f75
   - User: sr_read_write
   - Default Db
   - Sql Statement: SELECT COLUMN_NAME, DATA_TYPE, ORDINAL_POSITION, COLUMN_SIZE, DECIMAL_DIGITS, IS_NULLABLE, COLUMN_KEY, COLUMN_COMMENT FROM information_schema.COLUMNS WHERE TABLE_SCHEMA='lendtrade' AND TABLE_NAME='tr_tran_proc_db_sub';
   - Variables: parallel_fragment_exec_instance_num=8,max_parallel_scan_instance_num=-1,pipeline_dop=0,enable_adaptive_sink_dop=true,enable_runtime_adaptive_dop=false,runtime_profile_report_interval=10,resource_group=default_wg
   - NonDefaultSessionVariables: {"sql_mode_v2":{"defaultValue":32,"actualValue":2097184},"big_query_profile_threshold":{"defaultValue":"0s","actualValue":"30s"},"character_set_results":{"defaultValue":"utf8","actualValue":"NULL"},"parallel_fragment_exec_instance_num":{"defaultValue":1,"actualValue":8},"enable_adaptive_sink_dop":{"defaultValue":false,"actualValue":true},"enable_profile":{"defaultValue":false,"actualValue":true}}
   - Collect Profile Time: 2ms
   - IsProfileAsync: true
   Planner:
   - -- Parser[1] 0
   - -- Total[1] 0
   - -- Analyzer[1] 0
   - -- Lock[1] 0
   - -- AnalyzeDatabase[1] 0
   - -- AnalyzeTemporaryTable[1] 0
   - -- AnalyzeTable[1] 0
   - -- Transformer[1] 0
   - -- Optimizer[1] 0
   - -- MVPreprocess[1] 0
   - -- MVChooseCandidates[1] 0
   - -- MVGenerateMvPlan[1] 0
   - -- MVValidateMv[1] 0
   - -- MVProcessWithView[1] 0
   - -- MVTextRewrite[1] 0
   - -- RuleBaseOptimize[1] 0
   - -- CostBaseOptimize[1] 0
   - -- PhysicalRewrite[1] 0
   - -- PlanValidate[1] 0
   - -- InputDependenciesChecker[1] 0
   - -- TypeChecker[1] 0
   - -- CTEUniqueChecker[1] 0
   - -- ColumnReuseChecker[1] 0
   - -- ExecPlanBuild[1] 0
   - -- Pending[1] 0
   - -- Prepare[1] 0
   - -- Deploy[1] 23ms
   - -- DeployLockInternalTime[1] 23ms
   - -- DeploySerializeConcurrencyTime[1] 0
   - -- DeployStageByStageTime[3] 0
   - -- DeployWaitTime[3] 23ms
   - -- DeployAsyncSendTime[1] 0
   - DeployDataSize: 5803
   Reason:
   Execution:
   - Topology: {"rootId":1,"nodes":[{"id":1,"name":"PROJECT","properties":{"sinkIds":[],"displayMem":false},"children":[0]},{"id":0,"name":"SCHEMA_SCAN","properties":{"displayMem":false},"children":[]}]}
   - FrontendProfileMergeTime: 732.454us
   - QueryAllocatedMemoryUsage: 1015.680 KB
   - QueryCumulativeCpuTime: 11.169ms
   - QueryCumulativeNetworkTime: 0ns
   - QueryCumulativeOperatorTime: 240.565ms
   - QueryCumulativeScanTime: 229.352ms
   - QueryDeallocatedMemoryUsage: 840.773 KB
   - QueryExecutionWallTime: 257.617ms
   - QueryPeakMemoryUsagePerNode: 485.695 KB
   - QueryPeakScheduleTime: 42.480us
   - QuerySpillBytes: 0.000 B
   - QuerySumMemoryUsage: 485.695 KB
   - ResultDeliverTime: 0ns
   Fragment 0:
   - BackendAddresses: 172.17.172.252:9060
   - InstanceIds: 135d8852-62fd-11f0-b356-00163e164035
   - BackendNum: 1
   - BackendProfileMergeTime: 541.804us
   - FragmentInstancePrepareTime: 16.399ms
   - prepare-fragment-ctx: 922ns
   - prepare-pipeline-driver: 3.447ms
   - prepare-pipeline-driver-factory: 8.697ms
   - prepare-query-ctx: 3.826us
   - prepare-runtime-state: 4.247ms
   - InitialProcessDriverCount: 0
   - InitialProcessMem: 15.307 GB
   - InstanceAllocatedMemoryUsage: 1015.680 KB
   - InstanceDeallocatedMemoryUsage: 840.773 KB
   - InstanceNum: 1
   - InstancePeakMemoryUsage: 478.039 KB
   - JITCounter: 0
   - JITTotalCostTime: 0ns
   - QueryMemoryLimit: -1.000 B
   Pipeline (id=0):
   - isGroupExecution: false
   - ActiveTime: 11.183ms
   - BlockByInputEmpty: 2
   - BlockByOutputFull: 0
   - BlockByPrecondition: 0
   - DegreeOfParallelism: 1
   - DriverPrepareTime: 3.445ms
   - DriverTotalTime: 240.546ms
   - OverheadTime: 0ns
   - PeakDriverQueueSize: 0
   - PendingTime: 229.320ms
   - InputEmptyTime: 229.323ms
   - FirstInputEmptyTime: 229.245ms
   - FollowupInputEmptyTime: 78.368us
   - OutputFullTime: 0ns
   - PendingFinishTime: 0ns
   - PreconditionBlockTime: 0ns
   - ScheduleCount: 3
   - ScheduleTime: 42.480us
   - TotalDegreeOfParallelism: 1
   - YieldByLocalWait: 0
   - YieldByPreempt: 0
   - YieldByTimeLimit: 0
   RESULT_SINK (plan_node_id=-1):
   CommonMetrics:
   - IsFinalSink
   - CloseTime: 7.404us
   - OperatorAllocatedMemoryUsage: 19.320 KB
   - OperatorDeallocatedMemoryUsage: 20.281 KB
   - OperatorPeakMemoryUsage: 0.000 B
   - OperatorTotalTime: 5.905ms
   - PrepareTime: 3.384ms
   - PullChunkNum: 0
   - PullRowNum: 0
   - PullTotalTime: 0ns
   - PushChunkNum: 1
   - PushRowNum: 38
   - PushTotalTime: 5.897ms
   - SetFinishedTime: 30ns
   - SetFinishingTime: 60ns
   UniqueMetrics:
   result sink:
   - AppendChunkTime: 5.859ms
   - ResultRendTime: 33.463us
   - TupleConvertTime: 5.856ms
   - NumSentRows: 38
   CHUNK_ACCUMULATE (plan_node_id=-1):
   CommonMetrics:
   - IsSubordinate
   - CloseTime: 161ns
   - OperatorTotalTime: 1.133us
   - PrepareTime: 5.921us
   - PullChunkNum: 1
   - PullRowNum: 38
   - PullTotalTime: 280ns
   - PushChunkNum: 1
   - PushRowNum: 38
   - PushTotalTime: 613ns
   - SetFinishedTime: 50ns
   - SetFinishingTime: 29ns
   UniqueMetrics:
   PROJECT (plan_node_id=1):
   CommonMetrics:
   - CloseTime: 5.650us
   - OperatorAllocatedMemoryUsage: 1.016 KB
   - OperatorDeallocatedMemoryUsage: 416.000 B
   - OperatorPeakMemoryUsage: 640.000 B
   - OperatorTotalTime: 12.503us
   - PrepareTime: 7.204us
   - PullChunkNum: 1
   - PullRowNum: 38
   - PullTotalTime: 230ns
   - PushChunkNum: 1
   - PushRowNum: 38
   - PushTotalTime: 6.351us
   - RuntimeBloomFilterNum: 0
   - RuntimeInFilterNum: 0
   - SetFinishedTime: 141ns
   - SetFinishingTime: 131ns
   UniqueMetrics:
   - CommonSubExprComputeTime: 251ns
   - ExprComputeTime: 2.494us
   CHUNK_ACCUMULATE (plan_node_id=0):
   CommonMetrics:
   - IsSubordinate
   - CloseTime: 301ns
   - OperatorTotalTime: 2.636us
   - PrepareTime: 11.191us
   - PullChunkNum: 1
   - PullRowNum: 38
   - PullTotalTime: 191ns
   - PushChunkNum: 1
   - PushRowNum: 38
   - PushTotalTime: 1.823us
   - SetFinishedTime: 161ns
   - SetFinishingTime: 160ns
   UniqueMetrics:
   SCHEMA_SCAN (plan_node_id=0):
   CommonMetrics:
   - CloseTime: 64.822us
   - OperatorAllocatedMemoryUsage: 839.242 KB
   - OperatorDeallocatedMemoryUsage: 552.344 KB
   - OperatorPeakMemoryUsage: 307.672 KB
   - OperatorTotalTime: 5.292ms
   - PrepareTime: 16.070us
   - PullChunkNum: 1
   - PullRowNum: 38
   - PullTotalTime: 5.226ms
   - PushChunkNum: 0
   - PushRowNum: 0
   - PushTotalTime: 0ns
   - SetFinishedTime: 231ns
   - SetFinishingTime: 521ns
   UniqueMetrics:
   - MorselQueueType: fixed_morsel_queue
   - ChunkBufferCapacity: 64
   - DefaultChunkBufferCapacity: 64
   - FERPC: 211.509ms
   - FillChunk: 4.506ms
   - FilterTime: 12.890ms
   - IOTaskExecTime: 229.309ms
   - IOTaskWaitTime: 42.740us
   - MorselsCount: 1
   - PeakChunkBufferMemoryUsage: 250.513 KB
   - PeakChunkBufferSize: 1
   - PeakIOTasks: 1
   - PeakScanTaskQueueSize: 0
   - PrepareChunkSourceTime: 5.204ms
   - ScanTime: 229.352ms
   - SubmitTaskCount: 2
   - SubmitTaskTime: 8.015us
   - TabletCount: 1
   |

2. EXPLAIN ANALYZE

   +-------------------------------------------------------------------------------------------------------------------------------------+
   | Explain String |
   +-------------------------------------------------------------------------------------------------------------------------------------+
   | Summary |
   | QueryId: 6b148971-71e6-11f0-8cea-00163e39052a |
   | Version: 3.3.5-6d81f75 |
   | State: Finished |
   | TotalTime: 2s51ms |
   | ExecutionTime: 0ns [Scan: 0ns (NaN%), Network: 0ns (NaN%), ResultDeliverTime: 0ns (NaN%), ScheduleTime: 0ns (NaN%)] |
   | CollectProfileTime: 2s1ms |
   | FrontendProfileMergeTime: 361.325us |
   | QueryPeakMemoryUsage: ?, QueryAllocatedMemoryUsage: 0.000 B |
   | Top Most Time-consuming Nodes: |
   | 1. RESULT_SINK: 0ns (NaN%) |
   | 2. SCHEMA_SCAN (id=0) : 0ns (NaN%) |
   | Top Most Memory-consuming Nodes: |
   | NonDefaultVariables: |
   | big_query_profile_threshold: 0s -> 30s |
   | enable_adaptive_sink_dop: false -> true |
   | enable_async_profile: true -> false |
   | enable_profile: false -> true |
   | parallel_fragment_exec_instance_num: 1 -> 8 |
   | Fragment 0 |
   | │ BackendNum: 1 |
   | │ InstancePeakMemoryUsage: ?, InstanceAllocatedMemoryUsage: ? |
   | │ PrepareTime: ? |
   | │ MissingInstanceIds: 6b148971-71e6-11f0-8cea-00163e39052b |
   | └──RESULT_SINK |
   | │ SinkType: MYSQL_PROTOCAL |
   | └──SCHEMA_SCAN (id=0) |
   | Estimates: [row: 1, cpu: ?, memory: ?, network: ?, cost: 0.0] |
   | TotalTime: 0ns (NaN%) [CPUTime: ?] |
   | OutputRows: ? |
   | |
   +-------------------------------------------------------------------------------------------------------------------------------------+

3. ANALYZE PROFILE 展示的和explain analyze一样，但是指标比 explain analyze 更加丰富

   +-----------------------------------------------------------------------------------------------------------------------------------------------------------+
   | Explain String |
   +-----------------------------------------------------------------------------------------------------------------------------------------------------------+
   | Summary |
   | QueryId: 135d8852-62fd-11f0-b356-00163e164034 |
   | Version: 3.3.5-6d81f75 |
   | State: Finished |
   | TotalTime: 268ms |
   | ExecutionTime: 257.617ms [Scan: 229.352ms (89.03%), Network: 0ns (0.00%), ResultDeliverTime: 0ns (0.00%), ScheduleTime: 42.480us (0.02%)] |
   | CollectProfileTime: 2ms |
   | FrontendProfileMergeTime: 732.454us |
   | QueryPeakMemoryUsage: ?, QueryAllocatedMemoryUsage: 1015.680 KB |
   | Top Most Time-consuming Nodes: |
   | 1. SCHEMA_SCAN (id=0) : 234.646ms (97.54%) |
   | 2. RESULT_SINK: 5.906ms (2.46%) |
   | 3. PROJECT (id=1) : 12.503us (0.01%) |
   | Top Most Memory-consuming Nodes: |
   | NonDefaultVariables: |
   | big_query_profile_threshold: 0s -> 30s |
   | character_set_results: utf8 -> NULL |
   | enable_adaptive_sink_dop: false -> true |
   | enable_profile: false -> true |
   | parallel_fragment_exec_instance_num: 1 -> 8 |
   | sql_mode_v2: 32 -> 2097184 |
   | Fragment 0 |
   | │ BackendNum: 1 |
   | │ InstancePeakMemoryUsage: 478.038 KB, InstanceAllocatedMemoryUsage: 1015.680 KB |
   | │ PrepareTime: 16.399ms |
   | └──RESULT_SINK |
   | │ TotalTime: 5.906ms (2.46%) [CPUTime: 5.906ms] |
   | │ OutputRows: 38 |
   | │ SinkType: MYSQL_PROTOCAL |
   | └──PROJECT (id=1) |
   | │ Estimates: [row: ?, cpu: ?, memory: ?, network: ?, cost: ?] |
   | │ TotalTime: 12.503us (0.01%) [CPUTime: 12.503us] |
   | │ OutputRows: 38 |
   | │ Expression: [4: COLUMN_NAME, 5: ORDINAL_POSITION, 7: IS_NULLABLE, 8: DATA_TYPE, ...] |
   | └──SCHEMA_SCAN (id=0) |
   | Estimates: [row: 1, cpu: ?, memory: ?, network: ?, cost: 0.0] |
   | TotalTime: 234.646ms (97.54%) [CPUTime: 5.294ms, ScanTime: 229.352ms] |
   | OutputRows: 38 |
   | SubordinateOperators: |
   | CHUNK_ACCUMULATE |
   | Detail Timers: [ScanTime = IOTaskExecTime + IOTaskWaitTime] |
   | FERPC: 211.509ms |
   | IOTaskExecTime: 229.309ms |
   | IOTaskWaitTime: 42.740us |
   | |
   +-----------------------------------------------------------------------------------------------------------------------------------------------------------+

4. trace 展示的是某个局部，比如说 某个规则的耗时

   MySQL [(none)]> trace times SCHEDULE SELECT SCHEMA_NAME FROM INFORMATION_SCHEMA.SCHEMATA WHERE SCHEMA_NAME = 'fin_config';
   +--------------------------------------------------+
   | Explain String |
   +--------------------------------------------------+
   | 0ms|-- Parser[1] 0 |
   | 0ms|-- Total[1] 0 |
   | 0ms| -- Analyzer[1] 0 |
   | 0ms| -- Lock[1] 0 |
   | 0ms| -- AnalyzeDatabase[1] 0 |
   | 0ms| -- AnalyzeTemporaryTable[1] 0 |
   | 0ms| -- AnalyzeTable[1] 0 |
   | 0ms| -- Transformer[1] 0 |
   | 0ms| -- Optimizer[1] 0 |
   | 0ms| -- MVPreprocess[1] 0 |
   | 0ms| -- MVChooseCandidates[1] 0 |
   | 0ms| -- MVGenerateMvPlan[1] 0 |
   | 0ms| -- MVValidateMv[1] 0 |
   | 0ms| -- MVProcessWithView[1] 0 |
   | 0ms| -- MVTextRewrite[1] 0 |
   | 0ms| -- RuleBaseOptimize[1] 0 |
   | 0ms| -- CostBaseOptimize[1] 0 |
   | 1ms| -- PhysicalRewrite[1] 0 |
   | 1ms| -- PlanValidate[1] 0 |
   | 1ms| -- InputDependenciesChecker[1] 0 |
   | 1ms| -- TypeChecker[1] 0 |
   | 1ms| -- CTEUniqueChecker[1] 0 |
   | 1ms| -- ColumnReuseChecker[1] 0 |
   | 1ms| -- ExecPlanBuild[1] 0 |
   | Tracer Cost: 9us |
   +--------------------------------------------------+

   MySQL [(none)]> trace times OPTIMIZER SELECT SCHEMA_NAME FROM INFORMATION_SCHEMA.SCHEMATA WHERE SCHEMA_NAME = 'fin_config';
   +----------------------------------------------------------+
   | Explain String |
   +----------------------------------------------------------+
   | 0ms|-- Parser[1] 0 |
   | 0ms|-- Total[1] 0 |
   | 0ms| -- Analyzer[1] 0 |
   | 0ms| -- Lock[1] 0 |
   | 0ms| -- AnalyzeDatabase[1] 0 |
   | 0ms| -- AnalyzeTemporaryTable[1] 0 |
   | 0ms| -- AnalyzeTable[1] 0 |
   | 0ms| -- Transformer[1] 0 |
   | 0ms| -- Optimizer[1] 0 |
   | 0ms| -- MVPreprocess[1] 0 |
   | 0ms| -- MVChooseCandidates[1] 0 |
   | 0ms| -- MVGenerateMvPlan[1] 0 |
   | 0ms| -- MVValidateMv[1] 0 |
   | 0ms| -- MVProcessWithView[1] 0 |
   | 0ms| -- MVTextRewrite[1] 0 |
   | 0ms| -- RuleBaseOptimize[1] 0 |
   | 0ms| -- RewriteTreeTask[59] 0 |
   | 0ms| -- PushDownPredicateProjectRule[1] 0 |
   | 0ms| -- PushDownPredicateScanRule[1] 0 |
   | 0ms| -- MergeTwoProjectRule[2] 0 |
   | 0ms| -- PruneProjectColumnsRule[2] 0 |
   | 0ms| -- PruneScanColumnRule[2] 0 |
   | 0ms| -- PruneSubfieldRule[1] 0 |
   | 0ms| -- PruneProjectRule[2] 0 |
   | 0ms| -- MergeProjectWithChildRule[1] 0 |
   | 0ms| -- CostBaseOptimize[1] 0 |
   | 0ms| -- OptimizeGroupTask[1] 0 |
   | 0ms| -- OptimizeExpressionTask[1] 0 |
   | 0ms| -- DeriveStatsTask[1] 0 |
   | 0ms| -- ApplyRuleTask[1] 0 |
   | 0ms| -- SchemaScanImplementationRule[1] 0 |
   | 0ms| -- EnforceAndCostTask[1] 0 |
   | 0ms| -- PhysicalRewrite[1] 0 |
   | 1ms| -- PlanValidate[1] 0 |
   | 1ms| -- InputDependenciesChecker[1] 0 |
   | 1ms| -- TypeChecker[1] 0 |
   | 1ms| -- CTEUniqueChecker[1] 0 |
   | 1ms| -- ColumnReuseChecker[1] 0 |
   | 1ms| -- ExecPlanBuild[1] 0 |
   | Tracer Cost: 21us |
   +----------------------------------------------------------+

   其中这里面的[1] 数字代码该规则被应用的次数，如 PhysicalRewrite[1]则表示 改规则被应用了1次。