[
https://issues.apache.org/jira/browse/FLUME-1609?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13535765#comment-13535765
]
Juhani Connolly commented on FLUME-1609:
----------------------------------------
To clarify: in ganglia with this patch introduced, our system cpu usage would
shoot up to roughly double of the user cpu usage, and we would gradually see
the file channel get more full up. The difference was confirmed by testing with
both below revisions:
commit 8fb07e3c07c7b5fe37d629cee9100cab54a79819
Author: Hari Shreedharan <[email protected]>
Date: Tue Oct 16 14:36:44 2012 -0700
FLUME-1609. FileChannel detecting when the underlying file systems are full
could provide cleaner error recovery.
(Brock Noland via Hari Shreedharan)
commit 1dea8ec9d9359665f11b1e580fc4ea21e985d004
Author: Brock Noland <[email protected]>
Date: Tue Oct 16 14:27:40 2012 -0500
FLUME-1631: Retire hdfs.txnEventMax in HDFS sink
(Mike Percy via Brock Noland)
> FileChannel detecting when the underlying file systems are full could provide
> cleaner error recovery
> ----------------------------------------------------------------------------------------------------
>
> Key: FLUME-1609
> URL: https://issues.apache.org/jira/browse/FLUME-1609
> Project: Flume
> Issue Type: Improvement
> Components: Channel
> Affects Versions: v1.3.0
> Reporter: Brock Noland
> Assignee: Brock Noland
> Priority: Minor
> Fix For: v1.3.0
>
> Attachments: FLUME-1609-3.patch, FLUME-1609-4.patch
>
>
> At present if the file system fills up, we still try and write to it. We
> should change this such that if we know the fs is full we simply disable the
> log. This will result in less IO errors written to the log and many less
> extraneous log-* files created.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
