cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sylvain Lebresne (Commented) (JIRA)" <>
Subject [jira] [Commented] (CASSANDRA-2407) Compaction thread should try to empty a bucket before moving on
Date Thu, 17 Nov 2011 09:46:57 GMT


Sylvain Lebresne commented on CASSANDRA-2407:

* getNextBackgroundTask() computes the task based on all the sstables of the CF, not only
the one that are not being compacted. I believe it means that it will always return the same
sstables to compact as long as those are not finished to be compacted (the submitBackground
will then just say 'I can't mark as compacting so return' and we will essentially only have
one compaction at a time). A simple fix would probably be to add a getNonCompactingSSTables()
method to dataTracker and use that in getNextBackgroundTask.
* submitBackground doesn't resubmit a new compaction at the end, I think we still want to
do that.

Given this ticket does have the possibility to break stuffs, I'd be more confident delaying
it to 1.1.

> Compaction thread should try to empty a bucket before moving on
> ---------------------------------------------------------------
>                 Key: CASSANDRA-2407
>                 URL:
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Core
>            Reporter: Stu Hood
>            Assignee: Jonathan Ellis
>            Priority: Minor
>              Labels: compaction
>             Fix For: 1.1
>         Attachments: 2407-v2.txt, 2407-v3.txt, 2407.txt
> As suggested by Aaron Morton [(1)|],
a compaction thread should attempt to empty a bucket before moving on to a larger bucket.
This would change the submitMinorIfNeeded {{for}} loop into a while loop that regenerated
the buckets and started from the bottom after each successful compaction.

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:!default.jspa
For more information on JIRA, see:


View raw message