JBoss hangs after memory exceeds 160 000K

[lepekhine]

I use standard binary distribution of JBoss3.0alpha on NT Workstation 4.0. I have a simple application with 2 CMP EJB and one Web-component. After several database updates it hangs and I have to reboot JBoss to continue the work. Memory monitoring shows, that each update of EJB results in 1-2K memory allocation. When memory allcation grows to 160M JBoss hangs. When I try cancel it sometimes it doesn't response, sometimes notes about stack errors flushed on the screen. I have 391544K memory, so there are a lot of memory still available.
Detailed statistics is:
Memory allocated:
Initial 49320K
After JBoss started - 71536K
After IExplorer started - 77444K
Memory allocation increased about 1-2 K after each update
JBoss hangs on 156372K.
Can anybody help me?
And does anybody know when will be JBoss3.0beta released?
I have read in this forum that some bugs were fixed after 3.0alpha was released.

[lepekhine]

Me be this information will be usefull. In jboss.log the last lines are:
[2002-02-01 14:32:42,412,EnterpriseContextCachePolicy,DEBUG] Resized cache for bean Category: old capacity = 114, new capacity = 50
[2002-02-01 14:34:49,745,PoolGCThread,DEBUG] gc thread running gc
[2002-02-01 14:34:49,745,PoolGCThread,DEBUG] gc thread pool: NoTransDS isTimeToCleanup()true
[2002-02-01 14:34:49,745,NoTransDS,WARN] running gc for pool
[2002-02-01 14:34:49,745,PoolGCThread,DEBUG] gc thread pool: DefaultDS isTimeToCleanup()true
[2002-02-01 14:34:49,745,DefaultDS,WARN] running gc for pool
[2002-02-01 14:34:49,745,PoolGCThread,DEBUG] gc thread pool: JmsXA isTimeToCleanup()true
[2002-02-01 14:34:49,745,JmsXA,WARN] running gc for pool
[2002-02-01 14:34:49,745,PoolGCThread,DEBUG] gc thread waited for pools
[2002-02-01 14:34:49,745,PoolGCThread,DEBUG] gc thread delay: 600000
Don't know what can it mean.

[nick]

Funny, I have the some problem.

I am on Solaris 8.
Java (build 1.3.0)
512 Mb on a SUN Enterprise 450.

When jboss hits 160Mb Virtual memory (103Mb resident mem)
It hangs. You can no longer bind to jboss over rmi. You have to kill -9.

My app uses many-to-many, works ok, but over time, memory usage grows.

PS I am running my JAVA VM with -Xmx400m, which allows the VM to grow to 400Mb.

[lepekhine]

It's more funny that I still can't understand why it's happened. It seems to me that Sun "cmpcustomer" example (wich is very similar to my application) works fine!

[neilmcc]

Sounds just like a problem I'm currently wrestling with on RH 6.2. Once jBoss (2.4.4) gets to RSS 103 meg it hangs. Process sits there ok, but can't do jndi lookups or anything else for that matter. I'm running on a 300Mhz AMD box with 512 meg RAM (should be plenty).

I've taken a look at the JVM output using -Xhprof=sites, following advice from warjort but nothing has emerged as yet.

If anyone has anything further on this, would be really please to hear! ;)