[
https://issues.apache.org/jira/browse/CASSANDRA-1631?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12928325#action_12928325
]
Gary Dusbabek commented on CASSANDRA-1631:
------------------------------------------
This isn't going to be as simple as it was last time (shifting work on to the
CompactionManager and blocking). Part of the unload/init code, you guessed it,
shifts work of the CompactionManager and blocks (index creation to be precise).
I'm looking into ways of doing this that don't involve deadlock or refactoring
large pieces of code. Part of the problem is that we've started treating the
CompactionManager as a way to synchronize access to sstables. The problem with
that is that it is very course and the jobs submitted to it do a lot more
things that pure FS work.
> dropping column families and keyspaces races with compaction and flushing
> -------------------------------------------------------------------------
>
> Key: CASSANDRA-1631
> URL: https://issues.apache.org/jira/browse/CASSANDRA-1631
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Affects Versions: 0.7 beta 2
> Reporter: Gary Dusbabek
> Assignee: Gary Dusbabek
> Fix For: 0.7 beta 3
>
> Attachments: v1-0001-fix-drop-race-with-compaction.txt,
> v1-0002-fix-drop-race-with-flush.txt
>
>
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
