[
https://issues.apache.org/jira/browse/CASSANDRA-15045?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16797021#comment-16797021
]
Marcus Eriksson commented on CASSANDRA-15045:
---------------------------------------------
Pushed a new commit to the trunk branch to rename the variable
Also pushed the {{closeAll}} change to all branches and made sure we don't
error-log {{CompactionInterruptedException}}
bq. As usual, I want to double check its ok to change the semantics in each of
the versions the change is proposed in. Are their any cases where it should be
an option whether or not global compactions are cancelled? My initial
inclination is no but wanted to bring it up explicitly regardless.
It has been possible to cancel index summary redistributions before as well -
but only explicitly (like {{nodetool stop INDEX_SUMMARY}}) but not
automatically when starting anticompaction or truncating for example, so I
think it should be safe. My biggest worry was that since explicit stops like
above are much more rare than the automatic ones we would find some other issue
(like transaction not getting closed properly)
> Fix index summary redistribution compaction cancellation issues
> ---------------------------------------------------------------
>
> Key: CASSANDRA-15045
> URL: https://issues.apache.org/jira/browse/CASSANDRA-15045
> Project: Cassandra
> Issue Type: Bug
> Reporter: Marcus Eriksson
> Assignee: Marcus Eriksson
> Priority: Normal
> Fix For: 2.2.x, 3.0.x, 3.11.x, 4.x
>
>
> We can't cancel ongoing index summary redistributions currently due to
> {{CompactionInfo}} returning null for {{getTableMetadata/getCFMetaData}}
> [here|https://github.com/apache/cassandra/blob/67d613204fa4fb9584f11ec9886a0e7a0d622e92/src/java/org/apache/cassandra/db/compaction/CompactionManager.java#L1814]
> for index summary redistributions
> CASSANDRA-14935 also introduced a bug where we track the wrong sstables for
> index summary redistributions
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
