native_stderr max log size

[Sabina N]

I am not WebSphere admin, one of the developers for cycle changes for a rule-engine software running on WS6.
We have been having performance issues for quite some time and when I recently checked server logs I have noticed that size of native_stderr.log on the app server is almost 2.5 Gb. It's been containing info since Sept.01 this year. It seems rather unusual because admins at my previous job usually had a routine/scripts in place to do proper cut and archiving of logs.
It seems to be a bit out of control here. I see SystemErr and SystemOut being cut at every meg, but native_stderr keep growing over 2 gigs and nothing is done about it. Especially considering performance issues.

Can anybody confirm that such a huge size of native_stderr.log might impact web application performance and suggest some sort of policy/recommendation for truncating and archiving this log ( as well as native_stdout.log)?
Is there WS settings to arrange this or does it have to be done by script/manually?
Do we need to stop app server/WS server to truncate it?

PS. Not sure if it matters but the server elings to a cluster, second server on this cluster has native_stderr.log of 400Mb.

Thanks in advance,
Sabina

[Davanum Srinivas]

Sabina,

You may have to look at what gets dumped into the stderr.log as this log does not have options to be rotated as explained here:
http://javagongura.blogspot.com/2009/03/log-rotation.html

It's quite possible that you have "verbose garbage" enabled in the same admin console screen (shown in #3 figure in the url above). Or there is something else that is regularly dumping lots of information. You will have to track that down and fix the underlying problem.

Please also see this redbook that shows how do to problem determination. (info on native_stderr.log is on page 23)
http://www.redbooks.ibm.com/redbooks/pdfs/sg246798.pdf

thanks,
dims

[Sabina N]

Dims,

Thanks a lot for your reply. It's been tremendous help.
Now I can see why log is growing so hugh, we are dealing with rule engine with lots of generated code (25K total) and experienced OOMs quite often lately.
Looks like even there were dump files in the log directory (which I analyzed with IBMs tool), there was still dump written in this log file. I understand why this file cannot be rotated.
Redbook document says:
"After you resolve a problem, either delete or archive the diagnostic data that
you collected for the problem. This will prevent the old diagnostic data from
being confused with the new diagnostic data the next time that a problem
occurs."

I assume we can now safely rename this log as app server is running and the new log will be created ?

Thanks,
Sabina

[Akash Bharti]

native-stdout.log

i.e. back up the file and then clear the file.

Thanks
Akash