James, Since you told the context is rather huge, have you checked gc times? A long running full gc can block the machine completely resulting in the up/down behaviour from outside. GC options depend on JVM version I use:
export JAVA_OPTS="$JAVA_OPTS -XX:+DisableExplicitGC -verbose:GC -XX:+PrintGCDetails -XX:+PrintGCDateStamps -Xloggc:logs/gc.log " alternatively moskito would also show you avg response times and gcs ;-) regards Leon I agree. Long GC pauses can cause this type of issue. I have seen somewhat similar. Apache PING failed due to long GC pause and user got 503s. One question, do you see evidence of Tomcat restart in logs which correlates with alerts?
