[
https://issues.apache.org/jira/browse/HBASE-3160?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12926173#action_12926173
]
HBase Review Board commented on HBASE-3160:
-------------------------------------------
Message from: "Nicolas" <[email protected]>
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://review.cloudera.org/r/1103/#review1697
-----------------------------------------------------------
@Stack: Note that this is going into our 0.89 branch, so it should receive
thorough real-world testing. I would suggest the 0.90 branch. It's actually a
bug fix in addition to a feature. With the current compaction logic (pre
HBASE-2462), the CompactSplitThread can get overwhlemed without this. You
unconditionally compact at least 4 files to be aggressive in times of light
load. This is fine until your load picks up, you're congested, and writing too
many storefiles. Then, since the PriorityCompactionQueue is essentially
operating as a FIFO until the blocking limit, it doesn't make any informed
decisions about who to service next and waits until your puts freeze to be
intelligent...
- Nicolas
> Compactions: Use more intelligent priorities for PriorityCompactionQueue
> ------------------------------------------------------------------------
>
> Key: HBASE-3160
> URL: https://issues.apache.org/jira/browse/HBASE-3160
> Project: HBase
> Issue Type: Improvement
> Affects Versions: 0.89.20100924, 0.90.0
> Reporter: Nicolas Spiegelberg
> Assignee: Nicolas Spiegelberg
> Fix For: 0.90.0
>
>
> One of the problems with the current compaction queue is that we have a very
> low granularity on the importance of the various compactions in the queue.
> If a StoreFile count exceeds 15 files, only then do we bump via enum change.
> We should instead look into more intelligent, granular priority metrics for
> choosing the next compaction.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.