wangbo opened a new issue #2920: AlterJobV2 Pending Status Replay Fix Bug URL: https://github.com/apache/incubator-doris/issues/2920 **Bug Description** The TabletInvertedIndex isn't a Persistence class. In the Loading image process,Doris first load olaptable from image, Then using olaptable's index to create TabletInvertedIndex which means ```TabletInvertedIndex``` depends on olaptable metadata. In the``` AlterJobV2```'s implementation,first create TabletInvertedIndex before runpendingJob,then create olaptable index when pendingJob finished. In the replayPendingMethod,TabletInvertedIndex is created by ```AlterJobV2 ```'s metadata. But when pendingJob is checkpointed,replayPendingMethod won't be called. So in this case,if fe restarts,```TabletInvertedIndex``` won't be created; This may cause Fe think a tablet which has been created in be is invalid and delete that tablet in be,finally alter job failed. **To Reproduce** When alter job in pending status,restart fe twice to trigger checkpoint,then the job failed. For batch submit alter add rollup job, this is easier to reproduce. **Solution** Call replayPending method when load pending status AlterJobV2.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org For additional commands, e-mail: commits-h...@doris.apache.org
