ube fails during peak usages...

Hi all,

We're now in full production system and during the past months were encountering problems on our enterprise server where our ube's would fail daily during peak usages. A restart of the JDE service and the JAS (required since it would not authenticate properly if the JDE service is restarted only) would return the system to normal. Already applied the performance/memory esu's.

Were on all windows platform using 8.12/8.96G1, Oracle DB 10.2.0.1.0 and OAS 10.1.2.0.2. Enterprise and Jasserver have both 4GB memory with only about 25 users (10-15 concurrent users checked it using SAW)

I wouldn't mind restarting the server's on a nightly basis since i can schedule it but restarting it during a peak period of the day ughh.. it sucks.

Can anybody suggest what to look/check into... Is this a known problem? One thing i've noticed on our database is that our open cursors reached ~4500, is this normal for JDE?

thanks, hope somebody can help me go back to sleep...
KENTOY

this post from another site tells me that its a bug on the database...it exibited the same symptoms to our enterprise.

http://thetendjee.wordpress.com/2007/01/22/oracle-10202-sucks-on-windows/

have to cross my fingers on this.

Kentoy,

What service pack of windows 2003? SP2 by chance? We are on XE, SP23, but noticed that we cannot advance our operating system up to service pack 2 because long running UBEs fail. Shorter UBEs processed fine, but longer running UBEs like the package build process, crap out after 15 or 20 minutes. I don't see a resolution yet, other than backing the server down to SP1. The issue is the new MDAC that microsoft put out with Win 2k3 SP2.

Sopmething to look at.....

Gregg Larkin
Praxair North American System Admin
JDE CNC and Security, Websphere, Tidal, Princeton Softech

Hi Greg,

Nope, were still on SP1. I'm now waiting for a reply from oracle and our consultant on how we can access oracle's patchset from their metalink site (geez can't they made this a public download).

thanks.

Nah. They won't make it public. That's where they do the rest of their money (on top the licenses).

After patching to 10.2.0.3 in Japan, I've faced big performance issues due to the optimizer and bind variables (as specified on my article references in this thread).

The patch 10.2.0.3.5 corrects this.

During the 10.2.0.3.5 patching, I've faced a weird issue. 2 dll wouldn't copy and were considered in use even though they were not. Out of despair after 1 hour of maintenance (where it should be 20 minutes), i renamed those files (showing at the same time they were definitely not in use) and run the patch procedure again. That time it copied them...

If you are on 10.2.0.2 it's really a must to go for the 10.2.0.3

Hi Kentoy,

Did you manage to resolve this issue by upgrading oracle?

Thanks