MRP problems after Oracle upgrade...

[swhitmire]

We recently upgraded from Oracle 8.1.7.4 to 9.2.0.5 and we're having a problem with R3483.
We've only been able to successfully run the full regeneration once since the upgrade.

The first time it errored out we got this message in the jde.log:

59356 Wed Feb 16 07:27:03 2005 dbperfrq.c386
OCI0000178 - Unable to execute - SELECT IXTBM, IXKIT, IXMMCU, IXITM, IXCMCU, IXCOBY FROM PRODDTA.F3002 WHERE ( IXTBM = :KEY1 AND IXMMCU = :KEY2 AND IXKIT = :KEY3 AND IXCOBY = :KEY4 )

59356 Wed Feb 16 07:27:03 2005 dbperfrq.c393
OCI0000179 - Error - ORA-24371: data would not fit in current prefetch buffer

In the other runs, the only error in the log is that it ends with:

INFO: Done setting IPC Handle State structures to abandoned. Process exiting.

And I'm not sure that's an error and not just a result of the process dying.
(I saw the other thread about this message, but we're on SP22_W1 on an AIX box, so I don't think it's related.)

All of our other processes are running fine. Has anyone else experienced this problem?
Any ideas about what's causing it/how to solve it?

 

[cmanderson]

I see you are running Xe. We upgraded to 9.2.0.4 prior to the E1 8.9 upgrade and had the same issue with the prefetch buffer messages. I suggest changing your JDE.INI to use the libora80.so driver (ours was libora80.sl because we are running HP-UX). It will still work, at least it fixed the problem in our case. This may resolve some issues but it is not the permanent fix. Perhaps you should look at the latest service packs for Xe and determine if any fixes pertain to this problem?

 

[swhitmire]

Did you change all the data source definitions back to libora80 as well=3F=0D =0ADid you get the errors on lots of processes, or just some in particular=3F=0D =0A=0D=0A

 

[cmanderson]

Re: RE: MRP problems after Oracle upgrade...

You should change it in the data source configuration as well. Sorry, forgot to mention that. I wouldn't bother changing it for the fat clients and Citrix servers (if any), just the UNIX enterprise server. As for how many jobs were ending in error, it was a fairly small number but they were critical for the business. I recall a "known bug" being mentioned by GSC when I called them about the issue.