yujun777 opened a new pull request, #22861:
URL: https://github.com/apache/doris/pull/22861
## Proposed changes
Tablet sched will always failed for case below:
1. At the beginning, 4 BE: A, B, C, D; replica alloc num=3; three
replicas are on A, B, C;
2. User change BE B/C/D to tag:foo, also change table and partitions to
tag:foo;
3. Tablet scheduler will migrate replica from A => D.
4. Firstly, D clone a replica. During cloning, table loads data. After
cloning, there are 4 replicas:
a) A: version complete, tag mismatch;
b) B: version complete, tag match;
c) C: version complete, tag match;
d) D: version incomplete, tag match;
5. Tablet's aliveAndVersionComplete is ok(3 replicas: A,B,C), then it check
if replicas of each tag are enough , it will found tag:foo 's replica is not
engough, it has 2 replicas: B, C (cause the alloc map excludes version
incompleted D), so this tablet's health status is REPLICA_MISSING_FOR_TAG;
6. Scheduler will try to find another BE to clone a new replica, then it
will failed.
The PR is to fix this. For a tag, if the alive replicas are enough, but
version complete replicas are not enough, then the tablet's status should be
VERSION_INCOMPLETE.
## Further comments
If this is a relatively large or complex change, kick off the discussion at
[d...@doris.apache.org](mailto:d...@doris.apache.org) by explaining why you
chose the solution you did and what alternatives you considered, etc...
--
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.
To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org
For additional commands, e-mail: commits-h...@doris.apache.org
