import into job finished but no data exists when injecting one of tidb failure for 2mins

[Lily2025]

Bug Report

Please answer these questions before submitting your issue. Thanks!

1. Minimal reproduce step (Required)

1、import into user_data1 from 's3://brie/mvindex-10000000-csv/test.user_data1.*.csv?access-key=minioadmin&secret-access-key=minioadmin&endpoint=http://minio.pingcap.net:9001&force-path-style=true/' WITH skip_rows=1
2、 injecting one of tidb failure for 2mins

2. What did you expect to see? (Required)

import into job finished and checksum is correct

3. What did you see instead (Required)

cCrc32Checksum is 0, expected cCrc32Checksum is 3417923749

4. What is your TiDB version? (Required)

./tidb-server -V
Release Version: v7.4.0-alpha
Edition: Community
Git Commit Hash: 8f8c433
Git Branch: heads/refs/tags/v7.4.0-alpha
UTC Build Time: 2023-09-13 06:50:51
GoVersion: go1.21.1
Race Enabled: false
Check Table Before Drop: false
Store: unistore

[Lily2025]

/assign D3Hunter

[D3Hunter]

• during subtask running, tidb owner quite for a while, the subtask is marked as failed with cancelled error(guess from log), and dispatcher exit
• dispatcher restart, it sees the task as running, but it don't query failed subtask, so it think prev stage is finished and go into next stage directly
  

  tidb/disttask/framework/dispatcher/dispatcher.go

  Lines 237 to 243 in a7e3176

  	cnt, err := d.taskMgr.GetSubtaskInStatesCnt(d.Task.ID, proto.TaskStatePending, proto.TaskStateRunning)
  	if err != nil {
  	logutil.Logger(d.logCtx).Warn("check task failed", zap.Error(err))
  	return err
  	}
  	if cnt == 0 {

[D3Hunter]

fixed by #45623