[
https://issues.apache.org/jira/browse/HBASE-11120?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14005471#comment-14005471
]
Misty Stanley-Jones commented on HBASE-11120:
---------------------------------------------
a store can only have so many files, and that's called the blocking store file
count
so you cannot flush data if you have too many files, you have to wait for a
compaciton
and you cannot insert if the memstore is over some threshold
but then the compaction selection might not select a single file if none of
them are "good"
so if you get there, you're stuck
that's the accounting here:
boolean mayBeStuck = (candidateFiles.size() - filesCompacting.size() +
futureFiles)
>= storeConfigInfo.getBlockingFileCount();
Jean-Daniel Cryans
10:29 am
if our files we're supposed to compact, minus the ones we're already
compacting, plus 1 since those files will end up being a new store file, is
greater or equal to the blocking file count
then we force a compacting
compaction
> Update documentation about major compaction algorithm
> -----------------------------------------------------
>
> Key: HBASE-11120
> URL: https://issues.apache.org/jira/browse/HBASE-11120
> Project: HBase
> Issue Type: Bug
> Components: Compaction, documentation
> Affects Versions: 0.98.2
> Reporter: Misty Stanley-Jones
> Assignee: Misty Stanley-Jones
> Attachments: HBASE-11120.patch
>
>
> [14:20:38] <jdcryans> seems that there's
> http://hbase.apache.org/book.html#compaction and
> http://hbase.apache.org/book.html#managed.compactions
> [14:20:56] <jdcryans> the latter doesn't say much, except that you
> should manage them
> [14:21:44] <jdcryans> the former gives a good description of the
> _old_ selection algo
> [14:45:25] <jdcryans> this is the new selection algo since C5 /
> 0.96.0: https://issues.apache.org/jira/browse/HBASE-7842
--
This message was sent by Atlassian JIRA
(v6.2#6252)
