load phase of all mode may lost data when worker frequently starts/stops task

[lance6716]

Bug Report

Please answer these questions before submitting your issue. Thanks!

1. What did you do? If possible, provide a recipe for reproducing the error.

   will reproduce error later.

   This bug is caused by two hidden bugs

   a) cleanDumpFiles is wrongly triggered because it will check checkPoint.AllFinished(), but this is a memory cache of DB which may not initialized
   b) load unit didn't return error when there're missing SQL files which should exists according to checkpoint

   when worker is too quickly Close because of scheduling of network problems, worker may trigger a). In all mode, cleanDumpFiles will clean the SQL files (in full mode cleanDumpFiles will clean whole directory including dump metadata thus would cause an error). Then next time when it resume with the task, worker is continued on load unit because IsFresh will look into database rather than cache like checkPoint.AllFinished(). When load unit starts, it will trigger b) so finish load unit because of no files to load and goes to sync unit.

2. What did you expect to see?

3. What did you see instead?

4. Versions of the cluster

   • DM version (run dmctl -V or dm-worker -V or dm-master -V):

     (paste DM version here, and you must ensure versions of dmctl, DM-worker and DM-master are same)
     

[glkappe]

asktug link：https://asktug.com/t/topic/67791