It is definately Lucenes fieldCache making the trouble.
Restart your solr and monitor it with jvisualvm, especially OldGen heap.
When it gets to 100 percent filled use jmap to dump heap of your system.
Then use Eclipse Memory Analyzer http://www.eclipse.org/mat/ and
open the heap dump. You will see a pie chart and can easily identify
the largets consumer of your heap space.
Am 27.07.2011 09:02, schrieb Paul Libbrecht:
On curriki.org, our solr's Tomcat saturates memory after 2-4 weeks.
I am still investigating if I am accumulating something or something else is.
To check it, I am running a "query all, return num results" every minute to
measure the time it takes. It's generally when it meets a big GC that gives a timeout
that I start to worry. Memory then starts to be hogged but things get back to normal as
soon as the GC is out.
I had other tomcat servers with very long uptimes (more than 6 months) so I do
not think tomcat is guilty.
Currently I can only show the freememory of the system and what's in
solr-stats, but I do not know what to look at really...
paul
Le 27 juil. 2011 à 03:42, Bing Yu a écrit :
I find that, if I do not restart the master's tomcat for some days,
the load average will keep rising to a high level, solr become slow
and unstable, so I add a crontab to restart the tomcat everyday.
do you boys restart your tomcat ? and is there any way to avoid restart tomcat?
