[SOLVED] Server Restarts

bfolan · #1 (**permalink**) April 4th, 2007, 16:58

Is scheduled server restarts for good for OTM application? How often? Is this recommended? Would anyone know what type of issues this avoids?

We are having OTM performance issues and trying to find out if restarts will solve this. Any information on this would be greatly appreciated....

chrisplough · #2 (**permalink**) April 4th, 2007, 23:45

Most OTM clients have a scheduled restart in order to refresh the JVM and avoid and potential memory leaks that could affect performance. Prior to OTM/GC3 v5.0 - some of the higher volume clients were restarting daily in order to keep their instances stable and at peak performance. With v5.0 and especially v5.5 - most clients have been able to move towards longer restart cycles - usually weekly, bi-weekly or monthly. I usually recommend weekly or bi-weekly, whichever fits best into regular maintenance cycles.

One note - after a restart all of the OTM caches are empty and the JVM code optimizations (hotspot or otherwise) are cleared. As a result, OTM may be slow for a few transactions.

Hope this helps!

--Chris

runinord · #3 (**permalink**) April 24th, 2007, 13:46

We restart OTM every weekend due to our experience that if we do not the web server will fail very often after 8-9 days in the midle of the day. Furthermore, the BEA Weblogic also will start experiencing performance issues, e.g. sometimes a process will run a job for hours (until server is stoped and restarted) that takes the database to it's knees. We have had these issues open with OTM suport for 6-7 months but they have no solution except restarting OTM once a week. Hope that helped :-)

bfolan · #4 (**permalink**) April 24th, 2007, 14:49

Thanks for the response. That is the issue we are having. Very sporatic.

ianlo · #5 (**permalink**) April 26th, 2007, 06:05

This is also the same issue I am experiencing. However, my app tier JVM is the one that crashes - and there is a setting on Weblogic that will automatically restart the JVM after a crash.

The only problem with that is that if the JVM crashes and restarts in the middle of some operations (e.g. order releasing, planning etc) you will need to hunt down all the orders that failed and manually correct them. Very painful.

Better to reboot weekly.