Thanks Josh and Chris,
I have jira'ed this as:
http://opencast.jira.com/browse/MH-8205
Seeing others are effected I would suggest this is considered a blocker
for the upcoming release 1.3 release. Josh am I correct in reading that
patching any of the servelet filters would fix this for all sessions?
Regards
David
On 10/04/2011 06:47 AM, Josh Holtzman wrote:
We did use pax web in Matterhorn 1.0, or pre-1.0, but quickly moved to Felix's
own http bundle.
Unfortunately, there is no config in this http service implementation, see [1].
Adding a line like this to a servlet filter should fix it, though.
((HttpServletRequest) request).getSession().setMaxInactiveInterval(15 * 60);
Any filter will do (e.g. JsonpFilter, OrgFilter, SecurityFilter). If you want
to separate concerns, create a new filter to handle this specific need. I
could also see a filter that specifically invalidates sessions (or forces
spring security not to create them in the first place) from any request that
used digest auth. My gut feeling is that these are the sessions that are
causing the memory problems in the first place.
[1] https://issues.apache.org/jira/browse/FELIX-2687
Josh
On Oct 3, 2011, at 12:37 PM, Christopher Brooks wrote:
Hi David,
Memory dump was insightful, thanks for sharing. Seems that Jetty is
keeping sessions open for some reason; I see almost two gigs of
org.mortbay.jetty.servlety.HashSessionManager$Session classes open,
with no link back to our code. A quick question on #jetty revealed
this bug, which looks like the one I'm seeing in this dump file:
https://bugs.eclipse.org/bugs/show_bug.cgi?id=354297
In short, jetty defaults to session lengths of -1, which is never.
Resolution is to set session lengths to something like 30 mins.
(note: I don't really know how we're using jetty at all, it's a black
box to me, so this next bit is just from browsing the code to get a
better grip)
I think we use pax web to realise the http service, which uses jetty.
I think this means that you can configure jetty by changing the last
line of /felix/conf/bin/start_matterhorn.sh to read:
java $DEBUG_OPTS $GRAPHICS_OPTS $MAVEN_ARG $JAVA_OPTS
-Dorg.ops4j.pax.web.config.file=/opt/matterhorn/felix/conf/felix.xml
"$FELIX_FILEINSTALL_OPTS" "$PAX_CONFMAN_OPTS" $PAX_LOGGING_OPTS
$ECLIPSELINK_LOGGING_OPTS "$MATTERH
ORN_LOGGING_OPTS" "$UTIL_LOGGING_OPTS" -jar "$FELIX_HOME/bin/felix.jar"
"$FELIX_CACHE"
Then, create a file in /opt/matterhorn/felix/conf/felix.xml with some
magic configuration lines.
The magic configuration lines is where I'm not quite at I'm afraid.
Anyone want to share if
a) this is the best way to configure our web server?
b) does anyone know the magic juice to set the felix to time out
sessions?
Regards,
Chris
On Mon, 03 Oct 2011 16:37:50 +0200
David Horwitz<[email protected]> wrote:
Hi All,
Dropped the file here:
http://source.cet.uct.ac.za/opencastproject/memDump/HeapDump.hprof
Regards
D
On 10/03/2011 03:55 PM, Micah Sutton wrote:
I would also like to see the memory dump!
Micah
On Oct 3, 2011, at 5:07 AM, [email protected] wrote:
Hi david,
I'll bite on this if you can make the dump available...
chris
Quoting David Horwitz<[email protected]>:
Hi All
As mentioned in a thread last week I took a heap dump from our
production Matterhorn instance that had become unresponsive (100%
CPU but low load) and tried to run it through the eclipse memory
stand alone analyser. On my 32 bit desktop machine I can't
allocate enough heap to the analyser to be able to parse the
file. As this tool could successfully parse dumps from Sakai
instances with a very bad memory leak I would consider this
ominous.
I would gladly share the dump (3.5Gb) with others
Regards
David
_______________________________________________
Matterhorn mailing list
[email protected]
http://lists.opencastproject.org/mailman/listinfo/matterhorn
To unsubscribe please email
[email protected]
_______________________________________________
_______________________________________________
Matterhorn mailing list
[email protected]
http://lists.opencastproject.org/mailman/listinfo/matterhorn
To unsubscribe please email
[email protected]
_______________________________________________
_______________________________________________
Matterhorn mailing list
[email protected]
http://lists.opencastproject.org/mailman/listinfo/matterhorn
To unsubscribe please email
[email protected]
_______________________________________________
_______________________________________________
Matterhorn mailing list
[email protected]
http://lists.opencastproject.org/mailman/listinfo/matterhorn
To unsubscribe please email
[email protected]
_______________________________________________
--
Christopher Brooks, BSc, MSc
ARIES Laboratory, University of Saskatchewan
Web: http://www.cs.usask.ca/~cab938
Phone: 1.306.966.1442
Mail: Advanced Research in Intelligent Educational Systems Laboratory
Department of Computer Science
University of Saskatchewan
176 Thorvaldson Building
110 Science Place
Saskatoon, SK
S7N 5C9
_______________________________________________
Matterhorn mailing list
[email protected]
http://lists.opencastproject.org/mailman/listinfo/matterhorn
To unsubscribe please email
[email protected]
_______________________________________________
_______________________________________________
Matterhorn mailing list
[email protected]
http://lists.opencastproject.org/mailman/listinfo/matterhorn
To unsubscribe please email
[email protected]
_______________________________________________
_______________________________________________
Matterhorn mailing list
[email protected]
http://lists.opencastproject.org/mailman/listinfo/matterhorn
To unsubscribe please email
[email protected]
_______________________________________________
