PhanLe1010

@derekbit May I have question about case 7. It seems that this is the case in which replica successfully expanded but the engine crashed. In this case, how do we...

> Then, we should send both spec.size and status.currentSzie to the engine. Good idea

@w13915984028 is mentioning that this problem doesn't happen in Longhorn 1.3.0 However, I have not had time to test 1.3.0 yet.

Linking this PR to the corresponding issue https://github.com/longhorn/longhorn/issues/1473

Thanks @son-la for reporting It seems that the problem is caused by the introduction of the new longhorn-engine-manager and `longhorn-replica-manager` services, [link](https://github.com/longhorn/longhorn-manager/commit/f994a750c004934a11df4f6b815a8759822b577d) cc @keithalucas could you take a look to...

@son-la Can you dump out the firewalld config and the exact steps that you use to set up the env? We would like to reproduce the problem in our lab...

@derekbit to narrow down if the mismatch is caused by snapshot prune, could you try the same test with older Longhorn version (v1.2.4)?

> Before the prune, the volume-heads have different checksums. Is it because the prune operation does not have the lock protection? (cc @shuo-wu) Do you mean `After the prune, ....`...

Thank you for the quick confirmation. This is my wild thought. Can you give a comment on whether it makes sense @shuo-wu ? Volume head is actively being written to...

> We also need to add the chart to the upcoming 2.7 as well, but of course, it's not part of this OOB. cc @PhanLe1010 @rebeccazzzz Can we create a...