Three options:
Tweak it: Java Garbage Collectors (GC's), and the recommendations for
using them, have changed several times. I encourage you to find a more
recent article (e.g.,
http://www.oracle.com/technetwork/java/javase/gc-tuning-6-140523.html)
so you get the recommendations for the version of Java that you are
using. It's worth a try.
Fix it: The ideal approach is to identify why the problem is occurring,
not just deal with the consequences of the problem. Some of the issues
are programming related and can be solved or minimized via code changes,
but you have to have a clear idea of what the problem is. You can use
hprof to identify what lines of code are generating the most, persistent
objects. (There are other profilers.) And you can use -verbose:gc to
monitor the GC. I'm sure John Caron has investigated and will continue
to investigate thoroughly. But you are in a better position to
investigate the problems on your TDS.
Restart it: It is hard/impossible to prevent these problems from
occurring. The garbage collectors aren't perfect. Sometimes, increasing
MaxPermSize and Xmx and tweaking the GC settings just forestall the
inevitable. Sometimes, you just need to acquiesce and restart tomcat or
restart the computer. Although the big web services do as much as
possible to avoid problems like this, they rely on multiple servers and
load balancing so that a given computer can be restarted when needed.
Good luck.
On 6/1/2011 8:24 AM, Rich Signell wrote:
THREDDS folk,
Have folks played around with Garbage Collection tuning parameters
besides "-XX:MaxPermSize" to help tomcat/TDS permgen errors?
Our TDS tomcat is throwing "GC overhead limit exceeded" errors again.
I know on the Unidata TDS page
http://www.unidata.ucar.edu/projects/THREDDS/tech/tds4.2/tutorial/BasicTomcatAndTDSSecurity.html
it says the only thing you can do is increase -XX:MaxPermSize, which
only delays the problem, solved only by restarting tomcat.
But in this Garbage Collection tuning article
http://www.petefreitag.com/articles/gctuning/
they describe many different options, and I'm wondering whether folks
have tried using these, or whether they just don't do any good.
Thanks,
Rich
--
Sincerely,
Bob Simons
IT Specialist
Environmental Research Division
NOAA Southwest Fisheries Science Center
1352 Lighthouse Ave
Pacific Grove, CA 93950-2079
Phone: (831)658-3205
Fax: (831)648-8440
Email: bob.simons@xxxxxxxx
The contents of this message are mine personally and
do not necessarily reflect any position of the
Government or the National Oceanic and Atmospheric
Administration.
<>< <>< <>< <>< <>< <>< <>< <>< <><
