disttask: add more retryable error (#48033) (#48045)

ref #46258, close #48034
pingcap · Oct 30, 2023 · 3f2c2a2 · 3f2c2a2
1 parent bcd76f1
commit 3f2c2a2
Show file tree

Hide file tree

Showing 2 changed files with 17 additions and 1 deletion.
diff --git a/pkg/disttask/framework/scheduler/BUILD.bazel b/pkg/disttask/framework/scheduler/BUILD.bazel
@@ -20,10 +20,12 @@ go_library(
         "//pkg/disttask/framework/storage",
         "//pkg/domain/infosync",
         "//pkg/metrics",
+        "//pkg/parser/terror",
         "//pkg/resourcemanager/pool/spool",
         "//pkg/resourcemanager/util",
         "//pkg/util",
         "//pkg/util/backoff",
+        "//pkg/util/dbterror",
         "//pkg/util/logutil",
         "@com_github_pingcap_errors//:errors",
         "@com_github_pingcap_failpoint//:failpoint",

diff --git a/pkg/disttask/framework/scheduler/scheduler.go b/pkg/disttask/framework/scheduler/scheduler.go
@@ -29,7 +29,9 @@ import (
 	"github.com/pingcap/tidb/pkg/disttask/framework/storage"
 	"github.com/pingcap/tidb/pkg/domain/infosync"
 	"github.com/pingcap/tidb/pkg/metrics"
+	"github.com/pingcap/tidb/pkg/parser/terror"
 	"github.com/pingcap/tidb/pkg/util/backoff"
+	"github.com/pingcap/tidb/pkg/util/dbterror"
 	"github.com/pingcap/tidb/pkg/util/logutil"
 	"go.uber.org/zap"
 )
@@ -575,6 +577,18 @@ func (s *BaseScheduler) finishSubtaskAndUpdateState(ctx context.Context, subtask
 	metrics.IncDistTaskSubTaskCnt(subtask)
 }
 
+// TODO: abstract interface for each business to implement it.
+func isRetryableError(err error) bool {
+	originErr := errors.Cause(err)
+	if tErr, ok := originErr.(*terror.Error); ok {
+		sqlErr := terror.ToSQLError(tErr)
+		_, ok := dbterror.ReorgRetryableErrCodes[sqlErr.Code]
+		return ok
+	}
+	// can't retry Unknown err
+	return false
+}
+
 // markSubTaskCanceledOrFailed check the error type and decide the subtasks' state.
 // 1. Only cancel subtasks when meet ErrCancelSubtask.
 // 2. Only fail subtasks when meet non retryable error.
@@ -584,7 +598,7 @@ func (s *BaseScheduler) markSubTaskCanceledOrFailed(ctx context.Context, subtask
 		if ctx.Err() != nil && context.Cause(ctx) == ErrCancelSubtask {
 			logutil.Logger(s.logCtx).Warn("subtask canceled", zap.Error(err))
 			s.updateSubtaskStateAndError(subtask, proto.TaskStateCanceled, nil)
-		} else if common.IsRetryableError(err) {
+		} else if common.IsRetryableError(err) || isRetryableError(err) {
 			logutil.Logger(s.logCtx).Warn("met retryable error", zap.Error(err))
 		} else if errors.Cause(err) != context.Canceled {
 			logutil.Logger(s.logCtx).Warn("subtask failed", zap.Error(err))