[
https://issues.apache.org/jira/browse/CASSANDRA-13398?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15954767#comment-15954767
]
Stefan Podkowinski commented on CASSANDRA-13398:
------------------------------------------------
Post CASSANDRA-9143, {{CompactionManager.performAnticompaction()}} is now
expected to work in a different context since before, as we now do
anti-compactions before incremental repairs and not for full repairs at all
anymore. The only way the method is called is by passing UNREPAIRED_SSTABLE as
repairedAt value. At the same time we should only anti-compact from the
unrepaired set anyways (incr. repairs), but in the light of this filtering by
isRepaired() doesn't make sense anymore and is potentially dangerous. This of
course does also apply to the changes from CASSANDRA-13153 and not only to your
patch, Paulo.
I'd also suggest to look at the
mutatedRepairStatuses/mutatedRepairStatusToNotify/wasRepairedBefore
interactions again in this context. They should be either obsolete with Paulos
suggested patch, or dangerous when called with UNREPAIRED_SSTABLE as repairedAt.
/cc [~bdeggleston]
> Skip mutating repairedAt of already repaired sstables
> -----------------------------------------------------
>
> Key: CASSANDRA-13398
> URL: https://issues.apache.org/jira/browse/CASSANDRA-13398
> Project: Cassandra
> Issue Type: Bug
> Reporter: Paulo Motta
> Assignee: Paulo Motta
> Priority: Minor
>
> CASSANDRA-13153 skips anticompaction on sstables already repaired, but we are
> not skipping mutating repairedAt when the sstable is fully contained in the
> repaired range.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
