JVM utilising 100% CPU resources

Question

I have an application deployed to Oracle9iAS application server running on sunE280 with 2 processors and 4 higs ram. I have 1 oc4j process (JVM) serving my application. When there is heavy load on the system CPU utilisation peaks to 100% and the application stops responding. My oc4j processes (JVM) is running with 128m max heap size. So i increased the heap from default setting to min256m and Max512m sizes. I also increased the number of oc4j processes from 1 to 2.
At some point of time, both the oc4j processes utilise 48% of the CPU resouces each and then the application stops hanging. I did a vmstat and iostat on the system and below is o/p
procs memory page disk faults cpu
r b w swap free re mf pi po fr de sr m1 m2 m3 m4 in sy cs us sy id
0 0 0 4516376 2536400 25 76 150 0 0 0 0 0 0 1 0 349 99 1440 7 3 90
16 0 0 3945456 1787720 0 12 16 0 0 0 0 0 0 0 0 544 85221 2399 83 16 0
14 0 0 3945456 1787712 0 15 12 0 0 0 0 0 0 0 0 574 82593 2547 81 18 1
16 0 0 3944856 1787288 54 483 28 0 0 0 0 0 0 16 1 667 93353 2792 76 24 0
15 0 0 3940208 1787088 34 276 16 0 0 0 0 6 0 8 0 667 93823 2736 75 25 0
14 0 0 3945208 1787544 0 11 16 0 0 0 0 0 0 0 0 695 92164 3103 85 15 0
14 0 0 3945248 1787584 0 1 8 0 0 0 0 1 0 14 0 647 96116 2818 83 16 1
14 0 0 3945208 1787544 0 27 0 0 0 0 0 0 0 16 5 689 96923 2718 83 14 3
11 0 0 3945184 1787520 23 229 32 12 12 0 0 0 0 0 0 628 81816 2683 79 21 0
iostat 2 10
tty md1 md2 md3 md4 cpu
tin tout kps tps serv kps tps serv kps tps serv kps tps serv us sy wt id
0 20 3 0 42 0 0 28 4 1 15 2 0 20 7 3 2 88
0 0 0 0 0 0 0 0 0 0 0 0 0 0 83 16 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 81 18 0 1
0 0 0 0 0 0 0 0 8 16 8 0 1 21 76 24 0 0
0 0 48 6 56 0 0 0 4 9 6 0 1 38 75 25 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 85 14 0 0
0 0 1 1 20 0 0 0 65 12 24 0 0 0 82 16 0 1
0 0 0 0 0 0 0 0 41 19 20 18 5 22 83 14 0 3
Can someone provide me any clue as to why the JVM peaks in CPU utilisations and then stops responding?

Answer 1

Check Oracle's MetaLink (http:\\metalink.oracle.com) bug number 2218716. There is a problem with OC4J Containers and the FIXED_WAIT_SCHEME when the maximun number of connections is reached.
Just in case you don't have access to it, here is part of the document:
SUGGESTED FIX: Introduce a new cache property, CacheFixedWaitSleepTime, on OCCImpl (with setter/getter methods). In the while loop that yields when MAX+1 connection is requested, simply sleep for the number of milliseconds.
WORKAROUND: Use other cache schemes, (1) DYNAMIC_SCHEME (2) FIXED_RETURN_NULL_SCHEME
Hope the helps ...