Perplexing GC Time Growth

[Jason Vasquez]

During the latest test run, I did observe PU/PC for a bit -- although  
I did not notice in decrease in PU after a CMS (unfortunately i was  
only able to observe through one CMS).  Late in the test, here are the  
current values:  PC: 28824.0  PU: 16582.6    Doesn't actually seem  
like that large of a number for me.  PU was approximately 14157  
several hours earlier.

1) Yes, we do have a test setup where we can simulate full load and  
perform experiments, I would be more than happy to gather whatever  
data would be helpful.

2) Unfortunately, no official support contract at this point.  I'll  
see what can be done if it becomes absolutely necessary, but we're not  
a large company at this point :)

-jason




On Jan 3, 2008, at 13:10 , Y.S.Ramakrishna at Sun.COM wrote:

> Hi Jason -- thanks for the new data. I was wondering if you also
> got this data:-
>
>
>>> ... you might want to monitor
>>> the size of the perm gen (with and without the  
>>> CMSClassUnloadingEnabled
>>> option) by means of "jstat -gc ..." (which will show perm gen
>>> utilization under the "PU" column).
>
>
> And two follow-up questions:
>
> (1) do you have a test set-up on which some experiments could be  
> tried?
>    If so, i will be happy to suggest some follow-up data gathering
>    that might aid diagnosis.
>
> (2) do you have a Sun support contract that we could leverage to get
>    our support engineers looking at this issue? (If so, can you send
>    me either your contract #, or at least your company name so we can
>    dig up your support contract info).
>
> thanks.
> -- ramki