Hi Gurus !
We are experiencing some problems when using the async. I/O option in Oracle
8.1.6.3.1 or higher.
When patchkit 3 is applied and Oracle is patched to at least 8.1.6.3.1,
Oracle default uses async/direct I/O method to gain access to the disks
(like raw devices, but in AdvFS).
The servers crash when an Oracle filesystem (e.g. the archive disk) is 100%
full. We got a custom made patch to solve this problem. It appears to solve
the 100% full disk crash.
Yesterday we got another problem (panic) which caused the server to crash
several times yesterday :
_panic_string: 0xfffffc0000a384a0 = "live_dump: aio_run_completion:
ACQ_UNREF not set"
It also seems to be related to async. I/O (aio_run_completion)
We removed the custom made patch and disabled async i/o in Oracle (file:
init<SID>.ora) (parameter: disk_asynch_io = false).
Now the server has been running for 11½ hour, which is a new record.
Has anyone experienced similar problems ???
Any help is very much appreciated.....
With kind regards
Jan Hansen
Senior System Manager
LEGO
email : jan.hansen_at_europe.lego.com
Received on Fri Jul 20 2001 - 06:43:06 NZST