[
https://issues.apache.org/jira/browse/CASSANDRA-3181?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jonathan Ellis updated CASSANDRA-3181:
--------------------------------------
Attachment: 3181.txt
Attached patch does a couple things:
- Schedules a compaction submission for each CFS 5 minutes after startup. My
reasoning is, five minutes is (a) enough time for most caches to warm up under
load and (b) when it is not, at least it is enough time to reduce the
compaction i/o limit.
- removes the permanent check-for-compactions-every-3s task from leveled
compaction; I don't like spinning that up for no reason, when we already kick
off a check on each flush and end-of-compaction, which should be adequate.
(Every 3s for 1 CFS = every 0.0003s for 10K CFS.)
- makes Leveled getMaximal return a "normal, leveling" compaction, if any needs
to be done, allowing users of leveldb compaction to kick things off earlier
than 5m via "nodetool compact," if desired
> Compaction fails to occur
> -------------------------
>
> Key: CASSANDRA-3181
> URL: https://issues.apache.org/jira/browse/CASSANDRA-3181
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Affects Versions: 1.0.0
> Reporter: Brandon Williams
> Assignee: Benjamin Coverston
> Labels: compaction
> Fix For: 1.0.0
>
> Attachments: 3181.txt
>
>
> Compaction just stops running at some point. To repro, insert like 20M rows
> with a 1G heap and you'll get around 1k sstables. Restarting doesn't help,
> you have to invoke a major to get anything to happen.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
