> From: Bryan Wooten > Sent: Monday, August 31, 2015 12:58 PM > > So twice in the past few months CAS (3.5.2) has gotten really slow. A restart > of > the Tomcat servers makes the issue go away.
We've been using the Hazelcast backend for ages with no problems. The virtual machines running CAS have 2G and the following options: -Xms1025904k -Xmx1641446k -Xloggc:/var/log/tomcat-7/java-gc.log -verbose:gc -XX:+PrintGCDetails -XX:+PrintGCDateStamps -XX:+PrintGCTimeStamps -XX:+UseGCLogFileRotation -XX:NumberOfGCLogFiles=10 -XX:GCLogFileSize=100m -XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=/var/tmp/tomcat-7 -XX:+DisableExplicitGC -XX:+UseConcMarkSweepGC -XX:+UseParNewGC -XX:MaxGCPauseMillis=500 Right now over the summer we are only getting around 3-5k logins per day. We recently migrated to a new domain and I did not save the logs from our old CAS servers, so I'm not sure exactly what we were getting during a normal class session. If you are having lots of delay in garbage collection, you should be able to see it if you enable the garbage collection log file, it has various statistics including the time of the run, how much memory was shuffled around, and how long it took. We are currently evaluating migrating our CAS clients to the CAS protocol support in the shibboleth idp v3. Unicon is working on a Hazelcast storage backend for that too, although it is still in the development stage, not necessarily ready for production. -- Paul B. Henson | (909) 979-6361 | http://www.cpp.edu/~henson/ Operating Systems and Network Analyst | [email protected] California State Polytechnic University | Pomona CA 91768 -- You are currently subscribed to [email protected] as: [email protected] To unsubscribe, change settings or access archives, see http://www.ja-sig.org/wiki/display/JSG/cas-user
