[
https://issues.apache.org/jira/browse/QPID-2938?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12930709#action_12930709
]
Alan Conway commented on QPID-2938:
-----------------------------------
Created a patch to aggregate late/overrun values, see review
With the patch we still see warnings of many overrun events, but they overrun
by 2-3ms. There are no late events, and the test completes in < 30 seconds, is
comparable to a test with no TTL settings.
Patch is under review: https://reviews.apache.org/r/78/
> Logging excessive warnings in Timer code affects performance.
> -------------------------------------------------------------
>
> Key: QPID-2938
> URL: https://issues.apache.org/jira/browse/QPID-2938
> Project: Qpid
> Issue Type: Bug
> Components: C++ Broker
> Affects Versions: 0.6
> Reporter: Alan Conway
> Assignee: Alan Conway
>
> The Timer code logs a warning if a timer callback is started late or overruns
> the start time for the next callback.
> In cases where there are a lot of these warnings, the time taken to do the
> logging itself severly worsens the situation. You can provoke such a warnings
> storm by starting a broker in a cluster and running:
> qpid-cpp-benchmark -q4 --send-arg --ttl=10000
> The log shows timer events that are increaesingly late and overrun by up to
> 30 seconds, and the test is very slow to complete, I gave up after 2 minutes
> when the lateness was up to 90 seconds.
> This warning storm has also been seen with out clustering but with a store,
> see https://bugzilla.redhat.com/show_bug.cgi?id=629756
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project: http://qpid.apache.org
Use/Interact: mailto:[email protected]
