[
https://issues.apache.org/jira/browse/AMQ-4996?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13886365#comment-13886365
]
Jonathan Graham commented on AMQ-4996:
--------------------------------------
No high CPU tasks at that time. The system is sizing to cope with processing
10s of transactions per second but we're seeing this error even at very quiet
times (one transaction every 20-30 seconds) - ie producing a message every
20-30 seconds.
I didn't actually check the GC but there's very little going on so it's highly
unlikely that GC is causing such a long delay. Looking at my logs I can see
that around that time the average time we spent on a transaction (excluding
message production) is under 200ms with nothing at all above over 1.2s - if GC
was causing a pause I'd expect to see that impacting those timings as well.
> org.apache.activemq.transport.RequestTimedOutIOException thrown by
> producer.send() for no obvious reason. Message is actually sent.
> -----------------------------------------------------------------------------------------------------------------------------------
>
> Key: AMQ-4996
> URL: https://issues.apache.org/jira/browse/AMQ-4996
> Project: ActiveMQ
> Issue Type: Bug
> Components: Broker, JMS client
> Affects Versions: 5.5.1
> Environment: Redhat Linux
> Reporter: Jonathan Graham
>
> In production we are getting
> org.apache.activemq.transport.RequestTimedOutIOException exceptions thrown
> sometimes when we execute the send command. The timeout is set to 5 seconds
> and it blocks for 5 seconds and then throws the exception. The message is
> actually sent to the broker successfully (I don't know if it's sent
> immediately or if there was a delay before it gets there).
> We are communicating with the broker via TCP but it is on the localhost.
> This happens 'every so often' with no explanation that I can find. No errors
> appear in the broker logs (and no clues when configured to DEBUG level). The
> volume is very low; a message every 20-30 seconds or so, and they are
> immediately handled by a consumer - the broker has not run out of storage or
> memory.
> It seems to happen in bursts (e.g. it happened 4 times within 10 minutes, but
> not on every message, and then didn't happen again all day).
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
