statistics: add upper bound of retry for sync load (#52658)

statistics/handle/handle_hist.go

@@ -37,6 +37,9 @@
 	"golang.org/x/sync/singleflight"
 )
 
+// RetryCount is the max retry count for a sync load task.
+const RetryCount = 3
+
 type statsWrapper struct {
 	col *statistics.Column
 	idx *statistics.Index
@@ -56,6 +59,7 @@
 	TableItemID model.TableItemID
 	ToTimeout   time.Time
 	ResultCh    chan stmtctx.StatsLoadResult
+	Retry       int
 }
 
 // SendLoadRequests send neededColumns requests
@@ -217,6 +221,9 @@
 }
 
 // HandleOneTask handles last task if not nil, else handle a new task from chan, and return current task if fail somewhere.
+//   - If the task is handled successfully, return nil, nil.
+//   - If the task is timeout, return the task and nil. The caller should retry the timeout task without sleep.
+//   - If the task is failed, return the task, error. The caller should retry the timeout task with sleep.
 func (h *Handle) HandleOneTask(lastTask *NeededItemTask, readerCtx *StatsReaderContext, ctx sqlexec.RestrictedSQLExecutor, exit chan struct{}) (task *NeededItemTask, err error) {
 	defer func() {
 		// recover for each task, worker keeps working
@@ -236,53 +243,61 @@
 	} else {
 		task = lastTask
 	}
+	result := stmtctx.StatsLoadResult{Item: task.TableItemID}
 	resultChan := h.StatsLoad.Singleflight.DoChan(task.TableItemID.Key(), func() (any, error) {
-		return h.handleOneItemTask(task, readerCtx, ctx)
+		err := h.handleOneItemTask(task, readerCtx, ctx)
+		return nil, err
 	})
 	timeout := time.Until(task.ToTimeout)
 	select {
-	case result := <-resultChan:
-		if result.Err == nil {
-			slr := result.Val.(*stmtctx.StatsLoadResult)
-			if slr.Error != nil {
-				return task, slr.Error
-			}
-			task.ResultCh <- *slr
+	case sr := <-resultChan:
+		// sr.Val is always nil.
+		if sr.Err == nil {
+			task.ResultCh <- result
 			return nil, nil
 		}
-		return task, result.Err
+		if !isVaildForRetry(task) {
+			result.Error = sr.Err
+			task.ResultCh <- result
+			return nil, nil
+		}
+		return task, sr.Err
 	case <-time.After(timeout):
 		task.ToTimeout.Add(time.Duration(h.mu.ctx.GetSessionVars().StatsLoadSyncWait.Load()) * time.Microsecond)
 		return task, nil
 	}
 }
 
-func (h *Handle) handleOneItemTask(task *NeededItemTask, readerCtx *StatsReaderContext, ctx sqlexec.RestrictedSQLExecutor) (result *stmtctx.StatsLoadResult, err error) {
+func isVaildForRetry(task *NeededItemTask) bool {
+	task.Retry++
+	return task.Retry <= RetryCount
+}
+
+func (h *Handle) handleOneItemTask(task *NeededItemTask, readerCtx *StatsReaderContext, ctx sqlexec.RestrictedSQLExecutor) (err error) {
 	defer func() {
 		// recover for each task, worker keeps working
 		if r := recover(); r != nil {
 			logutil.BgLogger().Error("handleOneItemTask panicked", zap.Any("recover", r), zap.Stack("stack"))
 			err = errors.Errorf("stats loading panicked: %v", r)
 		}
 	}()
-	result = &stmtctx.StatsLoadResult{Item: task.TableItemID}
-	item := result.Item
+	item := task.TableItemID
 	oldCache := h.statsCache.Load().(statsCache)
 	tbl, ok := oldCache.Get(item.TableID)
 	if !ok {
-		return result, nil
+		return nil
 	}
 	wrapper := &statsWrapper{}
 	if item.IsIndex {
 		index, ok := tbl.Indices[item.ID]
 		if !ok || index.IsFullLoad() {
-			return result, nil
+			return nil
 		}
 		wrapper.idx = index
 	} else {
 		col, ok := tbl.Columns[item.ID]
 		if !ok || col.IsFullLoad() {
-			return result, nil
+			return nil
 		}
 		wrapper.col = col
 	}
@@ -292,8 +307,7 @@
 	needUpdate := false
 	wrapper, err = h.readStatsForOneItem(item, wrapper, readerCtx.reader)
 	if err != nil {
-		result.Error = err
-		return result, err
+		return err
 	}
 	if item.IsIndex {
 		if wrapper.idx != nil {
@@ -305,10 +319,10 @@
 		}
 	}
 	metrics.ReadStatsHistogram.Observe(float64(time.Since(t).Milliseconds()))
-	if needUpdate && h.updateCachedItem(item, wrapper.col, wrapper.idx) {
-		return result, nil
+	if needUpdate {
+		h.updateCachedItem(item, wrapper.col, wrapper.idx)
 	}
-	return nil, nil
+	return nil
 }
 
 func (h *Handle) loadFreshStatsReader(readerCtx *StatsReaderContext, ctx sqlexec.RestrictedSQLExecutor) {
@@ -493,12 +507,12 @@
 	oldCache := h.statsCache.Load().(statsCache)
 	tbl, ok := oldCache.Get(item.TableID)
 	if !ok {
-		return true
+		return false
 	}
 	if !item.IsIndex && colHist != nil {
 		c, ok := tbl.Columns[item.ID]
 		if !ok || c.IsFullLoad() {
-			return true
+			return false
 		}
 		tbl = tbl.Copy()
 		tbl.Columns[c.ID] = colHist

statistics/handle/handle_hist_test.go

@@ -209,6 +209,19 @@ func TestConcurrentLoadHistWithPanicAndFail(t *testing.T) {
 		require.Error(t, err1)
 		require.NotNil(t, task1)
 
+		select {
+		case <-stmtCtx1.StatsLoad.ResultCh:
+			t.Logf("stmtCtx1.ResultCh should not get anything")
+			t.FailNow()
+		case <-stmtCtx2.StatsLoad.ResultCh:
+			t.Logf("stmtCtx2.ResultCh should not get anything")
+			t.FailNow()
+		case <-task1.ResultCh:
+			t.Logf("task1.ResultCh should not get anything")
+			t.FailNow()
+		default:
+		}
+
 		require.NoError(t, failpoint.Disable(fp.failPath))
 		task3, err3 := h.HandleOneTask(task1, readerCtx, testKit.Session().(sqlexec.RestrictedSQLExecutor), exitCh)
 		require.NoError(t, err3)
@@ -231,3 +244,80 @@ func TestConcurrentLoadHistWithPanicAndFail(t *testing.T) {
 		require.Greater(t, hg.Len()+topn.Num(), 0)
 	}
 }
+
+func TestRetry(t *testing.T) {
+	originConfig := config.GetGlobalConfig()
+	newConfig := config.NewConfig()
+	newConfig.Performance.StatsLoadConcurrency = 0 // no worker to consume channel
+	config.StoreGlobalConfig(newConfig)
+	defer config.StoreGlobalConfig(originConfig)
+	store, dom := testkit.CreateMockStoreAndDomain(t)
+
+	testKit := testkit.NewTestKit(t, store)
+	testKit.MustExec("use test")
+	testKit.MustExec("drop table if exists t")
+	testKit.MustExec("set @@session.tidb_analyze_version=2")
+	testKit.MustExec("create table t(a int, b int, c int, primary key(a), key idx(b))")
+	testKit.MustExec("insert into t values (1,1,1),(2,2,2),(3,3,3)")
+
+	oriLease := dom.StatsHandle().Lease()
+	dom.StatsHandle().SetLease(1)
+	defer func() {
+		dom.StatsHandle().SetLease(oriLease)
+	}()
+	testKit.MustExec("analyze table t")
+
+	is := dom.InfoSchema()
+	tbl, err := is.TableByName(model.NewCIStr("test"), model.NewCIStr("t"))
+	require.NoError(t, err)
+	tableInfo := tbl.Meta()
+
+	h := dom.StatsHandle()
+
+	neededColumns := make([]model.TableItemID, 1)
+	neededColumns[0] = model.TableItemID{TableID: tableInfo.ID, ID: tableInfo.Columns[2].ID, IsIndex: false}
+	timeout := time.Nanosecond * mathutil.MaxInt
+
+	// clear statsCache
+	h.Clear()
+	require.NoError(t, dom.StatsHandle().Update(is))
+
+	// no stats at beginning
+	stat := h.GetTableStats(tableInfo)
+	c, ok := stat.Columns[tableInfo.Columns[2].ID]
+	require.True(t, !ok || (c.Histogram.Len()+c.TopN.Num() == 0))
+
+	stmtCtx1 := &stmtctx.StatementContext{}
+	h.SendLoadRequests(stmtCtx1, neededColumns, timeout)
+	stmtCtx2 := &stmtctx.StatementContext{}
+	h.SendLoadRequests(stmtCtx2, neededColumns, timeout)
+
+	exitCh := make(chan struct{})
+	require.NoError(t, failpoint.Enable("github.com/pingcap/tidb/statistics/handle/mockReadStatsForOneFail", "return(true)"))
+	var (
+		task1 *handle.NeededItemTask
+		err1  error
+	)
+	readerCtx := &handle.StatsReaderContext{}
+	for i := 0; i < handle.RetryCount; i++ {
+		task1, err1 = h.HandleOneTask(task1, readerCtx, testKit.Session().(sqlexec.RestrictedSQLExecutor), exitCh)
+		require.Error(t, err1)
+		require.NotNil(t, task1)
+		select {
+		case <-task1.ResultCh:
+			t.Logf("task1.ResultCh should not get nothing")
+			t.FailNow()
+		default:
+		}
+	}
+	result, err1 := h.HandleOneTask(task1, readerCtx, testKit.Session().(sqlexec.RestrictedSQLExecutor), exitCh)
+	require.NoError(t, err1)
+	require.Nil(t, result)
+	select {
+	case <-task1.ResultCh:
+	default:
+		t.Logf("task1.ResultCh should get nothing")
+		t.FailNow()
+	}
+	require.NoError(t, failpoint.Disable("github.com/pingcap/tidb/statistics/handle/mockReadStatsForOneFail"))
+}