james-server-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Bernd Fondermann" <bernd.fonderm...@googlemail.com>
Subject Re: Confirmed but unidentified memory leak in RC2
Date Mon, 11 Sep 2006 15:05:57 GMT
On 9/11/06, Vincenzo Gianferrari Pini
<vincenzo.gianferraripini@praxis.it> wrote:
>
>
> Stefano Bagnara wrote:
>
> >Noel J. Bergman wrote:
> >
> >
> >>Unfortunately, I once again confirm (this time using the embedded memstat
> >>check in RemoteManager) that we are losing about 2MB per day.  And it does
> >>seem related to the days, not the amount of mail.
> >>
> >>
> >
> >Imho the memstat showing 2MB more every day in an "uncontrolled"
> >environment is not a clear sign of a leak. Maybe you simply have
> >increase in traffic or anything else.. Imho this will be "confirmed"
> >when we'll be able to reproduce it in a different environment.
> >
> >
> >
> I agree, 2MB is "nothing" in a day.
>
> >
> >
> >>I am still prepared to release, but we will have to document this issue.  I
> >>would also like for everyone to check for this in their environment.  Once a
> >>day, run memstat -gc in the RemoteManager, and document the available heap
> >>after the garbage collection.
> >>
> >>      --- Noel
> >>
> >>
> >
> >Imo it does not make sense to document the issue until we are able to
> >reproduce it somewhere else and we will understand WHERE is the leak.
> >
> >
> Again, I agree
>

We need a reproducible case until we have a bug.

And we have to agree upon what load James must be able to handle.

We can always generate load to produce OOMs, we can always argue where
"leakage" becomes pathetic. I proposed some time ago to define a
"bottom line" the server must be able to handle.

What strikes me with the current report, is that it seems to be the
case that an environment previously working just fine over a rather
long time without problems is now experiencing problems after very
short uptime. Each release of James should at least as performant than
the previous. (And thats seems to be generally the case with the
current 2.3, as your reports indicate)

So how do we reproduce this beast? Do we have a "no load" reproduction
raising OOMs?

  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