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 Tue, 12 Sep 2006 11:50:35 GMT
On 9/12/06, Noel J. Bergman <noel@devtech.com> wrote:
> Bernd Fondermann wrote:
>
> > Noel J. Bergman wrote:
>
> > > It is quite possible that it is an old bug that is surfacing now because
> > > other bugs got out of the way.  There were old bugs that caused me to
> have
> > > to cycle JAMES 2.2 on a daily basis to keep it from throwing OOM.
>
> > are you saying, you upgraded directly from "2.2/daily restart" to
> > "2.3xx/memory leak".
>
> Yes.
>
> > has there been a "2.3yy/memory fine" in between?
>
> There hasn't been a "memory fine" in 2.x.  In production there have always
> been some "leaks", such as the fact that file-based repositories could
> monotonically increase their memory requirements, e.g., when used to record
> spam or error messages.  We've made progress to eliminate them, and I'm just
> disappointed that we have one now that is being difficult to track.

My understanding was that you were experiencing a degression.

> Last night I got a page that the server was unavailable.  I found a half
> dozen OOM in phoenix.console, and the SMTP threads were all non-responsive,
> but I could still telnet into RemoteManager, and saw that the heap was down
> to 8MB, but I could force the GC, and it came back with 100MB free.  JAMES
> didn't recover and required a re-start.  I did a kill -3 first, but the heap
> profile doesn't show anything obviously out of whack.

Maybe the OOM is only a sympton for other resources going out of
control, e.g. threads.
J5/JMX/JConsole would be able to track that.

  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