Re: WebSphere/HATS Odyssey
- From: Paul Dineen <dineenpj@xxxxxxxxx>
- Date: Tue, 4 Oct 2005 09:43:29 -0500
On Tue, 4 Oct 2005 09:46:55 -0400, Patrick.Falcone@xxxxxxxxxxxx wrote:
>I was in one of the many meetings that still go on, although our WAS 5
>conversion is basically on hold pending an upgrade, and heard that WAS V5
>is not 64 bit compliant. I guess I was asleep at the wheel to some extent
>but this poses some concern from my perspective when we do get an upgrade.
>
>Basically my concern is that we have quite a few HATS related
>applications, 26 and counting and currently running on NT, that will run
>as part of one of our cell groups as part of the WAS V5 conversion. With
>that said and with the 31 bit limitation my concerns are with populating
>the application servants with fewer applications than we may have hoped
>for. This will mean more servants with less applications per servant which
>means more storage and more CPU. I confess that I have not RTFM, can
>anyone confirm that WAS tasks have a 2 gb address space limit. WHAT, is
>the cart before the horse again!
>
>----------------------------------------------------------------------
Hello Patrick,
I believe we have realized the 2GB limit or close to it at WAS 5.1,
unfortunately in a production environment. Unfortunately, it sounds like
the application is not ready for scalability to additional WAS address
spaces.
We're working with IBM (Sev 1) on what is described as a 'memory leak in
JAVA'. I see the problem manifest as the working set to the WebSphere
address space climbs continuously to about 1.8GB. At that point, we
terminate.
IBM support has been asked about 64-bit in regards to our problem and
responds: "As for the question about our 2GB limit-this is something that
we are hoping to correct in a future release of the product".
Just to take a shot anyway, we tried setting a MEMLIMIT on the WAS EXEC
statement, but the message at startup still shows:
+BBOO0302I REGION REQUESTED = 0K 731
ACTUAL BELOW/ABOVE LINE LIMIT = 7M / 1842M
ABOVE BAR FREE/ALLOC ADDR = 0 / 0
It looks like WebSphere is 'preparing for 64-bit' at our 5.1 level by
displaying above the bar info, but not exploiting unless we're missing
something. This is our initial attempt at 64-bit. We're at z/OS 1.6, JDK
1.4 and all prereqs to utilize zAAP on a z/890.
Not to be a 'name dropper', but Mark (yes, he rocks) Zelden described
earlier this month that he was setting a MEMLIMIT for z/OS 1.6, WAS 5.1
and DB2 V8. Mark, any insight on whether you have WAS 5.1 exploiting 64-
bit? This would be a great patch for a shop with a production problem.
Hope this info helps you and we also can gain more insight.
Paul Dineen
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to listserv@xxxxxxxxxxx with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
.
- Follow-Ups:
- Re: WebSphere/HATS Odyssey
- From: Patrick . Falcone
- Re: WebSphere/HATS Odyssey
- Prev by Date: Re: SOC2 Problem in user defined macro
- Next by Date: Re: Amazing DFSORT behavior
- Previous by thread: WebSphere/HATS Odyssey
- Next by thread: Re: WebSphere/HATS Odyssey
- Index(es):
Relevant Pages
|
