At 11:08 AM 02/03/05, you wrote:
Is anyone experiencing sporadic core dumps and have any suggestions to eliminate them?
I'm running Unidata 6.0.3 on HPUX 11.11. The hardware is an RP5470 (2 cpus, 6GB memory). The application software is Datatel's Colleague and Benefactor. I have only indirect support from IBM for Unidata (I have to call Datatel for support).
Datatel first indicates that it may be old compiled code. I find it all, and recompile and catalog (if needed) all our custom. Recompiled every I-descriptor. Still get dumps.
Datatel then indicates that it might be level 2 overflow in the files. I've memresize-d the files and dicts and now udtmon shows no level-2 activity. Still got dumps, but not as frequently.
Datatel then says it might be index overflows. I recreate the indexes with larger node sizes to eliminate overflow reads/writes. Now there is none. Still get the occasional dump.
Datatel then says it must be hardware. There are no errors logged. I know hardware monitoring is working since I was getting single-bit errors reported 6 months ago and replaced affected memory, and a raid drive failure 9 months ago.
Last Thursday, I have a rash of failures of unidata in a 20 minute period. I've got dumps of some of the processes. Not all udt processes were affected. No hardware events recorded. Of the dumps I looked at the top of the stack was: U_free_varmem(4012EFC0) from U_freevar+0xA4 (or similar). I did reboot the machine, no problems.
Guide is run nightly on all accounts and their files using Datatel's WUFA process. There are no broken files reported and no overflow2 statuses.
Last night I find a powerpoint presentation "Problem Determination Techniques for Diagnosing UniData Database and System Issues" by Wally Terhune at the IBM DB2 Info. Mgmt. Technical Conference, Sept 20-24, 2004. There is a page that says "Working with Unix core files...UniData engineering likely at fault...". I am beginning to believe that I'm just spinning my wheels.
Having come from Prime Information, piopen, and universe environments where a dump/failure meant you have a problem, these occasional dumps are frustrating.
Does anyone have any ideas?
________________________________________________
Mike Cannady
Computing Services
Campbell University
PO Box 164
Buies Creek, NC 27506
[EMAIL PROTECTED]
(910) 893-1204
+----------------------------------------------+
| There is no place like 127.0.0.1 (aka ::1)! |
+----------------------------------------------+ ------- u2-users mailing list u2-users@listserver.u2ug.org To unsubscribe please visit http://listserver.u2ug.org/
Kathy Tymoczko
Information Technology Services
Wabash College, Crawfordsville, Indiana 47933
[EMAIL PROTECTED] 765-361-6440 -------
u2-users mailing list
u2-users@listserver.u2ug.org
To unsubscribe please visit http://listserver.u2ug.org/
