[ 
https://issues.apache.org/jira/browse/AXIS2-6009?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17405173#comment-17405173
 ] 

Eoghan O'Hare commented on AXIS2-6009:
--------------------------------------

Hi Robert,

The soft limit (-S) for root I had increased to 10000 (from 1024), the hard 
limit (-H) for open files for root was 262144 (that was the default already for 
the system).

I changed the /etc/security/limits.conf for root to unlimited;

root soft nofile unlimited

The ulimit -S -n now also shows 262144 (the same as the hard limit). 

After this change I got the same behavior for the load test.

Based on the RHEL article -XX:+MaxFDLimit is supposed to be set by default. To 
confirm, here is the default java VM argument from the system for MaxFDLimit.

[root@pmdpriepm52 VPAppIntfClient]# java -XX:+PrintFlagsFinal | grep MaxFD
 bool MaxFDLimit = true \{product}

I explicitly set -XX:+MaxFDLimit in the JAVA_OPTS of the application just to be 
sure. No difference.

When running the load test there was nothing outputted to /var/log/messages or 
/var/log/secure unfortunately.

Is there any reason other than ulimits that could prevent the number of http 
connections increasing above 5? I am not seeing any exceptions in the apache 
logs or anything.

I tried to go through the httpclient code on github, got to a certain point, 
but it was not easy to follow around the connection manager and leasing 
connections from the pool etc.

Is there any additional apache httpclient tracing you think I could enable that 
would help debug?

Regards,

Eoghan

 

> ServiceClient::createClient intermittently under load blocks for 5 seconds
> --------------------------------------------------------------------------
>
>                 Key: AXIS2-6009
>                 URL: https://issues.apache.org/jira/browse/AXIS2-6009
>             Project: Axis2
>          Issue Type: Bug
>    Affects Versions: 1.7.9
>            Reporter: Eoghan O'Hare
>            Priority: Major
>         Attachments: ExceptionStackTrace.txt, axis2.xml, axis2client.7z, 
> axis2client_httpclient4.log
>
>
> Using the axis2-kernel-1.7.9.jar.
> Our application is using automatically generated WS stubs for asynchronous 
> SOAP web services over https. 
> ServiceClient::createClient is called for every asynchronous web service 
> request.
> Callbacks are registered using OperationClient::setCallback().
> The non blocking web service request is sent by the API 
> OperationClient::execute(false). 
> The option isUseSeparateListener is set to false.
> The default number of http connections are created, this is two from looking 
> at pcap traces.
> Under low load rates, 20 requests per second, 1000 total requests, 
> intermittently the call to ServiceClient::createClient blocks the calling 
> thread for 5 seconds. Seems to be just over 5 seconds every time it occurs 
> (between 5000ms and 5030ms), so it sounds like some timeout.
> The blocking seems to occur when a response to a previous web service is 
> received at a similar time.
> After the 5 second blocking, an Exception is thrown from axis2 to our 
> applications error callback. Exception stack trace is attached.
> I have tried to enable axis2 logging on the client side however there was 
> nothing obvious in it from my point of view. Attaching the log. In the log 
> the 5 second delay is between;
> 06:47:50:574
> 06:47:55:579
> If you require additional logging to be enabled to debug please let me know.
> Or if you need any further information, please let me know.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscr...@axis.apache.org
For additional commands, e-mail: java-dev-h...@axis.apache.org

Reply via email to