You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "Jonathan Ellis (JIRA)" <ji...@apache.org> on 2010/01/05 18:36:54 UTC
[jira] Created: (CASSANDRA-670) commitlog sync seems to be affected
by i/o on other devices
commitlog sync seems to be affected by i/o on other devices
-----------------------------------------------------------
Key: CASSANDRA-670
URL: https://issues.apache.org/jira/browse/CASSANDRA-670
Project: Cassandra
Issue Type: Bug
Reporter: Jonathan Ellis
Priority: Minor
doing some testing on a system with commitlog on a separate device + xfs filesystem, syncing shouldn't have to wait for IO from the data filesystem, but we're seeing CL sync times up to almost 5s when the data fs is extra busy (adding bonnie++ to normal cassandra activity helps reproduce this). CASSANDRA-668 band-aided this a bit but it would be good to figure out what's going on here: is it our code at fault? the JVM's? the linux kernel's?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (CASSANDRA-670) commitlog sync seems to be affected
by i/o on other devices
Posted by "Jonathan Ellis (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-670?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Jonathan Ellis updated CASSANDRA-670:
-------------------------------------
Attachment: 670.txt
patch that logs when any sync takes longer than 1s (should be obvious from stack trace which ones are from commitlog)
> commitlog sync seems to be affected by i/o on other devices
> -----------------------------------------------------------
>
> Key: CASSANDRA-670
> URL: https://issues.apache.org/jira/browse/CASSANDRA-670
> Project: Cassandra
> Issue Type: Bug
> Reporter: Jonathan Ellis
> Priority: Minor
> Attachments: 670.txt
>
>
> doing some testing on a system with commitlog on a separate device + xfs filesystem, syncing shouldn't have to wait for IO from the data filesystem, but we're seeing CL sync times up to almost 5s when the data fs is extra busy (adding bonnie++ to normal cassandra activity helps reproduce this). CASSANDRA-668 band-aided this a bit but it would be good to figure out what's going on here: is it our code at fault? the JVM's? the linux kernel's?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (CASSANDRA-670) commitlog sync seems to be
affected by i/o on other devices
Posted by "Jonathan Ellis (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-670?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12803450#action_12803450 ]
Jonathan Ellis commented on CASSANDRA-670:
------------------------------------------
CASSANDRA-724 mitigates this more for periodic sync configuration.
> commitlog sync seems to be affected by i/o on other devices
> -----------------------------------------------------------
>
> Key: CASSANDRA-670
> URL: https://issues.apache.org/jira/browse/CASSANDRA-670
> Project: Cassandra
> Issue Type: Bug
> Reporter: Jonathan Ellis
> Priority: Minor
> Attachments: 670.txt
>
>
> doing some testing on a system with commitlog on a separate device + xfs filesystem, syncing shouldn't have to wait for IO from the data filesystem, but we're seeing CL sync times up to almost 5s when the data fs is extra busy (adding bonnie++ to normal cassandra activity helps reproduce this). CASSANDRA-668 band-aided this a bit but it would be good to figure out what's going on here: is it our code at fault? the JVM's? the linux kernel's?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.