james-server-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Bernd Fondermann <bf_...@brainlounge.de>
Subject Re: [jira] Commented: (JAMES-512) OutOfMemoryError -- object not released.
Date Fri, 02 Jun 2006 11:48:13 GMT
Stefano Bagnara wrote:
> My patch should has nothing to do with db repositories. So I think this 
> doesn't matter.

quick recap of my conclusions around JAMES-512:

a specific problem with file-repos was reported and your fix was applied.

after the fix, the problem symptoms (OOMs) don't completely go away, 
they only come appearant later.

so, the fix seems to help, but presumably doesn't make the problem go 
away. (another possibility is, that it's a completely different problem 
with similar symptoms.)

at this time, the question should be posed: is it still only a file-repo 
problem, or a more general problem?

that's the whole point of comparing file-repo with non-file-repo under 
the same conditions and the result is: db-repo has similar 
problems/symptoms. (and that's why it matters to repeat the tests with 
db repos.)

knowing this, JAMES-512 could be closed because it reports only a very 
specific memory leak.

next, we could open a new JIRA or could conclude that handling 500 
mails/min really needs more memory than 64MB.

> Instead I would like to know wether this is derby or mysql or anything 
> else?

> I know many connector/j uses big buffers and maybe they also have leak 
> problems. Derby instead have big caches, maybe it is a leak or maybe we 
> simply have to tune derby parameters (like we already did)

I used default derby.

   Bernd

---------------------------------------------------------------------
To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org


Mime
View raw message