cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jonathan Ellis (JIRA)" <j...@apache.org>
Subject [jira] Commented: (CASSANDRA-1404) Allow 'nodetool cleanup' to run when disk is more than 50% full
Date Thu, 23 Dec 2010 01:14:01 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-1404?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12974460#action_12974460
] 

Jonathan Ellis commented on CASSANDRA-1404:
-------------------------------------------

That's a reasonable improvement to make; I'd also like to make cleanup operate on one sstable
at a time, and _just_ clean out unneeded rows, rather than doing clean + major-compaction-merging.
 That way you can continue to rely on minor compactions if that fits your needs.

> Allow 'nodetool cleanup' to run when disk is more than 50% full
> ---------------------------------------------------------------
>
>                 Key: CASSANDRA-1404
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-1404
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Core
>            Reporter: Stu Hood
>            Priority: Minor
>             Fix For: 0.7.1
>
>
> Compaction sums the sizes of all ranges to estimate the output size in order to pick
a target drive. Anti-compactions directly drop ranges from sstables, and the volume of data
dropped can easily be calculated using the SSTable index. Using this knowledge in our estimate
in ColumnFamilyStore.getExpectedCompactedFileSize would allow 'nodetool cleanup' to run when
less than 50% of disk is available, in the case where ranges are being dropped (post node-move).

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message