Oracle 8.1.6, patches and relinking questions

[Stephanie Nicholls]

Hi Folks

I'm in the process of trying to track down a bizarre, infrequent, _apparently_ 'random' Oracle database instability problem.

Scenario: HPUX11.00 on a L3000, Oracle 8.1.6.3 production database approximately 100Gb in size 600+ concurrent users, 24 x 7.

Approximately 10 times in the last 7 months the db has gone 'unstable' and been unable to process user requests, in the trace files we've had snp processes dying and being respawned frantically. About 4 of those times the CKPT or PMON process has terminated the database.

2 of the last 3 incidents have been on days where disk changes have taken place. First time I had created 2 additional raw volumes in the same vg as the database, and second time (2 weeks later) the DBA added those raw volumes to the tablespaces.

Prior to june 2002 the db had been running 'perfectly' on the same server, the only change of significance was a large Patch bundle was applied in mid May.

Having seen a bit of discussion about relinking oracle after applying patches Im trying to work out when this activity should be done, and how can I tell if it was or should of been done after the last patch bundle.

Additionally after tracking through every Oracle prereq patch for 8.1.6, Ive discovered we do not have the Auxiliary-Opt product installed nor related patches, is this only required for development environments or production as well?

Any ideas or thoughts welcome

Cheers

Steph

[Ravi_8]

Hi steph

Applying patches shouldn't cause any problem to database. If the relinking was not proper then database wouldn't start first of all.

another thing i can suggest you is moving from 11.0 to 11i

[David_246]

Hi Steph,

Please look at the process table when your system is getting unstable. I had a similar problem after apply-ing a complete patch-bundle. If vhand shows up when your system is about to die, you'd better install patch : PHKL_27212 . This realy solved all my issues.

Regs David

[Brian Crabtree]

How big is your log files currently? Also, what errors and trace files were associated with the crash?

Can you list out the values for the following parameters:

log_buffer
db_block_buffers
shared_pool_size
log_checkpoint_interval
db_block_max_dirty_target

Thanks,

Brian

[Steven E. Protter]

Couple things to look at.

The December 2002 patch set. I includes a lot of patches that deal with Oracle Issues.

There are a few supplmental required patches that have been added int he last few months. Here is a list. You will need to check these to make sure there are not new patches from HP superseding these.

HP-UX 11.00 ===========
16. init.ora parameter PGA_AGGREGATE_TARGET may cause HP-UX panic First fixed by:
PHKL_25188 :
PHKL_28180 :
PHKL_26800.
15. Various Issues with HP-UX signals Current Revision of Patch:
PHKL_28152
PHKL_26059.

14. Large page size performance issues Current Revision of Patch:
PHKL_21180 : s700_800 11.00 Improved largepage performance for data PHKL_21180.

13. Various Hangs/Panics/Corruptions Enables o/s striping in shared logical volumes Enable OPS on FC60 array (allow 'vgchange -a s') PHKL_27178

12. Oracle Uses Single CPU on Multi-CPU Box After Applying PHNE_20008
PHNE_20094 : s700_800 11.00 PHNE_27821
PHNE_25385
PHNE_20008
PHNE_22566
PHNE_27651.

11. Possible Hang if 2 Processes Write to a Memory Mapped File First fixed by:
PHKL_19800: s700_800 11.00 fix VxFS unmount hang & MMF, sync panics More Current Revision of Patch: PHKL_27980
PHKL_24027.

10. Deadlock/Hang Upon After Adding an Extra CPU First fixed by:
PHKL_19647: s700_800 11.00 Fix rare TLB multiprocessor hang More Current Revision of Patch: PHKL_20915:
PHKL_20915.

9. Various problems with multi threaded Pro*Lang / Java Programs Current Revision of Patch: PHCO_26960
PHCO_23792.
8. ORA-7445 and Core Dump Generating Oracle Stack Trace on 64 bit Oracle RDBMS First fixed by:
PHSS_16849
PHSS_27469
PHSS_25718.

7. Space not freed up after a Vxfs file is removed. Introduced with: PHKL_17205 Fixed by: PHKL_17869 : 11.00 PM/VM/UFS/async/scsi/io/DMAPI/JFS/perf cumulative patch More Current Revision of PHKL_18543

6. Core dumps and other various problems mounting Oracle CD's using pfs_mount Fixed by: PHCO_14112
PHCO_16438

5. Various Network Problems Fixed by: PHNE_16645
PHNE_27886
PHNE_26771

4. various issues with ld during relinking Fixed by:
PHSS_15316
PHSS_26559
PHSS_24303

3. libc cumulative header patch Current Revision of Patch:
PHCO_26111 : s700_800 11.00 libc cumulative header file patch
PHCO_23963
2. libc cumulative patch Current Revision of Patch: PHCO_27731
PHCO_24148
PHCO_20555
1. Motif Problems Current Revision of Patch: PHSS_27230
PHSS_25447
PHSS_27232
PHSS_25879


Warning, install any of these other patches with the -x reinstall=false option.

reinstalling PHKL_18543 will cause you to TRASH your system. You will be unable to build kernels and lots and lots of patches will be ruined. HP's recommended recovery from that mistake is to reinstall the OS.

I know how to get around it but its very hard.

P

Also, make sure java is current along with patches.

1.2 or 1.3 are the only versions certified for Oracle

http://www.hp.com/go/java

P