On 16.1.2013 9:45, Lubos Kosco wrote:
On 16.1.2013 2:32, Devlin, Alex wrote:
Hello,
We're using OpenGrok 11.1 with a Perforce repository and Tomcat 6.
Unfortunately, we also have a memory leak in the application. It
sporadically crashes with very little warning beforehand; the
server's swap memory usage gradually starts rising about 3 hours
before it uses 100% of the system memory. We applied the bugfix
located at http://defect.opensolaris.org/bz/show_bug.cgi?id=19215
<http://defect.opensolaris.org/bz/show_bug.cgi?id=19215> but it did
not resolve the issue...
Get to 0.11.1 , I think no reindex needed, it should contain some fixes
we use 0.11.1 inhouse (not with perforce, but with 4 other scms) and
don't see any of this afaik (Vlada, any comments from your side?)
d'oh
sorry, ignore, misread, you are on 0.11.1 already ... :( , then start
with below, so jhat, jmap, mat and lets see what leaks ... it can be
just the perforce history analyzer, since we really don't see anything
on our server (but then we don't use perforce scm ... )
sorry
L
Right now we think it's some kind of thread leaking issue. When
Tomcat's shutdown.sh is called, there will often be several dozen
messages in the log files similar to the following:
Message was "SEVERE: The web application [/source] created a
ThreadLocal with key of type
[org.opensolaris.opengrok.configuration.RuntimeEnvironment$1] (value
[org.opensolaris.opengrok.configuration.RuntimeEnvironment$1@20a594d3])
and a value of type
[org.opensolaris.opengrok.configuration.Configuration] (value
[org.opensolaris.opengrok.configuration.Configuration@56cdb963]) but
failed to remove it when the web application was stopped. This is
very likely to create a memory leak."
check on tomcat docs, but this doesn't really point to leak, it just
means we persist objects for configuration (which are rather small,
unless Jens changed something more there)
I'd really start with 0.11.1, if still the problem will be there(I am
aware about another "leak", but not such as you describe and it's
caused by bug in lucene it seems, hope Alan will clarify here)
then I'd move to jmap, jhat, etc. (I think eclipse has mat or
something like that to analyze leaks) and do some more investigation,
but I vaguely remember we fixed something like this in 0.11.1
- the other thing is to give the webapp container more JVM memory, I
think the Oracle internal big opengrok service runs with 4G for the
container
(I hope we can lower these insane memory requirements by improving
analyzers in next version and stop caching whole files in memory)
xing the fingers
L
Any and all advice would be much appreciated. We're attempting to
locate the leak, but have not yet had much luck.
Thanks,
-Alex
_______________________________________________
opengrok-discuss mailing list
opengrok-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/opengrok-discuss
_______________________________________________
opengrok-discuss mailing list
opengrok-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/opengrok-discuss
_______________________________________________
opengrok-discuss mailing list
opengrok-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/opengrok-discuss
