[ 
https://issues.apache.org/jira/browse/QPID-6944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15064761#comment-15064761
 ] 

Keith Wall commented on QPID-6944:
----------------------------------

Hi Mark

I took a lool at qpid.log and the other files. Nothing stands out to me as 
unusual.

Can I suggest two experiments to hopefully help us narrow in on the problem?

# As I mentioned above, in v6.0.0, the AMQP IO was rewritten from scratch to 
use Java NIO.  I would like to rule out a problem in this area.  Can you apply 
SelectorThread.patch to the 6.0.0 source and repeat the reproduction, with 
logging set to DEBUG whilst the client is run for a period for, say, 6-8 
spikes.  Capture the qpid.log and a corresponding CPU usage graph from 
jvisualvm.  The patch just adds a couple of extra log statements.

# Can you repeat the reproduction a second time with logging reverted to normal 
settings?  This time use the Sampler in jvisualvm to capture the method 
hotspots.  (Go into the jvisualvm, connect to the Broker's VM, go to the Sample 
tab, and click CPU.  Capture a screen dump after another 6-8 spikes.  No Broker 
changes required.

What is the nature of the host hardware?


> High CPU use sending message
> ----------------------------
>
>                 Key: QPID-6944
>                 URL: https://issues.apache.org/jira/browse/QPID-6944
>             Project: Qpid
>          Issue Type: Bug
>          Components: Java Broker
>    Affects Versions: qpid-java-6.0
>         Environment: Red Hat Enterprise Linux Server release 6.7 (Santiago)
> OpenJDK Runtime Environment (build 1.8.0_65-b17)
> OpenJDK 64-Bit Server VM (build 25.65-b01, mixed mode)
>            Reporter: Mark Soderquist
>            Assignee: Keith Wall
>         Attachments: broker.acl, config.json, default.json, enabledebug.png, 
> qpid.0.32.png, qpid.6.0.png, qpid.cpu.png, qpid.cpu.png, qpid.log.zip, 
> qpid6944.zip, threaddump-1449856663270.tdump, virtualhosts.xml
>
>
> The org.apache.qpid.server.transport.SelectorThread is causing live lock. 
> While the application still functions it causes the CPU to run at nearly 
> 100%. See attached thread dump.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to