topic Re: Oracle10g mmnl errors in Operating System - HP-UX

Oracle10g mmnl errors

Greg Hall — Mon, 14 Nov 2005 16:52:07 GMT

Hi,

my client has a new oracle 10g r2 database on HPUX 11.11. In the database bdump dir this is a .trc file dbname_mmnl_13961.trc. Which contains:

Dump file /u00/app/oracle/admin/PRDCLRTY/bdump/prdclrty_mmnl_13961.trc
Oracle Database 10g Enterprise Edition Release 10.2.0.1.0 - 64bit Production
With the Partitioning, OLAP and Data Mining options
ORACLE_HOME = /u00/app/oracle/product/10.2.0.1.0
System name: HP-UX
Node name: akldwhp1
Release: B.11.11
Version: U
Machine: 9000/800
Instance name: PRDCLRTY
Redo thread mounted by this instance: 0
Oracle process number: 0
Unix process pid: 13961, image: oracle@akldwhp1

File '/dev/async' not present : errno=2
*** 2005-11-14 04:19:34.764
*** SERVICE NAME:(SYS$BACKGROUND) 2005-11-14 04:19:34.764
*** SESSION ID:(325.1) 2005-11-14 04:19:34.763
ksugetosstat failed: op = pstat_getprocessor, location = slsgetactive ()
*** 2005-11-14 04:19:47.136
ksugetosstat failed: op = pstat_getprocessor, location = slsgetactive ()
*** 2005-11-14 04:20:02.285
ksugetosstat failed: op = pstat_getprocessor, location = slsgetactive ()
*** 2005-11-14 04:20:16.423

- they are working on the /dev/async problem.
- are the ksugetosstat failures related to the /dev/async problem, or are they something else? I suspect the latter as these msgs are being produced every few secs since 0419 Monday.
Any clues would be welcome, I'm a bit lost.

thanks, Greg

Re: Oracle10g mmnl errors

Warren_9 — Mon, 14 Nov 2005 22:03:49 GMT

hi,

similar reported in metalink.oracle

https://metalink.oracle.com/metalink/plsql/ml2_documents.showDocument?p_database_id=FOR&p_id=613649.993

This problem happened on oracle8 and it come out again on the 10gR2....!?

GOOD LUCK!

Re: Oracle10g mmnl errors

Yogeeraj_1 — Mon, 14 Nov 2005 23:08:05 GMT

hi greg!

also refer to metalink notes:
Note.60692.1 ALERT HP-UX Tracefiles reference /dev/async when async i/o not configured

Note.1053060.6 TRACE FILE BEING GENERATED WITH EACH DBWR PROCESS

As the previous post say this error was first reported in Oracle 8i whereby the workaround was:

WORKAROUND
==========
It is possible to workaround the problem by the following means;

If the "/dev/async" device does not exists, then create it.

As the root user:

/sbin/mknod /dev/async c 101 0x0
/usr/bin/chown oracle:dba /dev/async
/usr/bin/chmod 000 /dev/async

The permissions on /dev/async of 000 ensure that it is not possible to use the async driver unintentionally.

As the oracle user, modify the "init.ora" file so that the parameters:

disk_asynch_io = false
tape_asynch_io = false

I would also suggest that you create a TAR in Metalink for a more proper followup. Most probably they already have a patch for that... please check

kind regards
yogeeraj

Re: Oracle10g mmnl errors

Greg Hall — Mon, 14 Nov 2005 23:27:42 GMT

thanks people. I'm aware of the fix for the dev/async problem. This is with the Oracle support people.

I'm more interested in the "ksugetosstat failed" error. Is this related to the /dev/async problem? Could be, in which case I should ignore them until the dev/async issue is resolved.

Greg

Re: Oracle10g mmnl errors

Eric Antunes — Tue, 15 Nov 2005 06:49:05 GMT

Hi Greg,

The only way to know if the ksugetosstat error message is related to /dev/async file is to - assuming you use file systems and not raw devices - modify the init.ora parameter disk_asynch_io to false (as Yogeeraj said), restart the database and see if those trace files continue to be created.

Best Regards,

Eric Antunes

Re: Oracle10g mmnl errors

Dennis Murphy_3 — Mon, 10 Jul 2006 09:13:07 GMT

Can I ask a question?

Does this particular system have any iCOD CPU's?

I have a 48-way Superdome with 16 iCOD CPU's, and I'm getting the same trace errors about "ksugetosstat failed".

The only guess I have is that since I have 16 iCOD CPU's, Oracle can't access them. If you query the number of CPUs in the system (for example, from a print_manifest), I show 64 CPU's, but only 48 are active.

I -think- it's related to iCOD, but I'm not 100% sure yet.

Re: Oracle10g mmnl errors

Adam Mullen — Fri, 21 Jul 2006 15:24:38 GMT

I'm also seeing this error on HPUX 11.11 running Oracle 10g r2. I don't understand why one would want to set disk_asynch_io to false though. Please explain this. Also, was a tar opened, if so, could you please provide the ID #.

Thanks, Adam

Re: Oracle10g mmnl errors

Mike Tufariello — Mon, 03 Jan 2011 18:00:03 GMT

I noticed the dates on this thread and see they are older.

We just moved to Oracle 10g r2 on a PA-RISC running 11.11 OS. We too are now seeing hundreds of .trc files being created. Does the above thread still apply as a vaild solution ? Or has more research and other methods been developed to address this issue ?

Thanks