Crash Analysis Help

[Steven Chen_1]

Hi,

I would like to have HP+Oracle guru to analyse the suddent Oracle crash last Friday. Your any comments are very apprciated.

Os is hpux 10.20, Oracle is V732. Physical Mem is 1G, Physical: 1048564 Kbytes, lockable: 778008 Kbytes, available: 895064 Kbytes. Oracle sga is 400M.

The crash was regarded either a memory false or process overflow that exceed memory. In any case, the reco process was not able to recover. Db was dead then. I had to recover all from the point of crash.

It would be interested for all viewers to learn from this instance. For your expertise, I would post the kernel configuration for your comment.

Thanks a lot first,

Steven

[Mladen Despic]

Steven,

Have you monitored swap utilization?
What does 'swapinfo -tam' show?
In any case, if you have enough disk for about 2Gb of swap device, I would suggest disabling pseudo-swap
(change the kernel parameter 'swapmem_on' to 0)

Regards ... Mladen

[Alessandro Bocchino]

You could disable dynamic buffer cache
by setting bufpages to a value different from 0, e.g. 10000 (40megs) to avoid possible allocation/deallocation of buffer pages.

Verify you have the latest HPUX kernel patches
installed.

Consider updating Oracle RDBMS to the latest
patch level, there are many issues related
to excessive memory allocation by snapshot
or shadow processes and processes not releasing
memory.