async io error message

[David S. Bell]

We upgraded Oracle to 10g on the production and standby databases. We are using home grown scripts not DataGusard to update the standby. Now when we replay the archive logs to update the standby database we get the following message. "OS async I/O limit 128 is lower than recovery batch 642". The job is now taking more than twice as long to apply the logs. We are running on HP-UX 11.11 and don't have the async drivers in the kernel on either Production or Standby servers. Are there kernel parameters that need updating? Do I need to install async in the kernel even if I'm not using raw disk?
We are asking Oracle the same questions.
Thank you in advance.

[Steven E. Protter]

Shalom,

There may be a problem in your asynchronous i/o setup.

See this document:
http://blog.chinaunix.net/u/12960/showart_446628.html

http://hpux-tips.blogspot.com/2006/01/how-to-enable-asynchronous-io-disk-hp.html

http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=103995

Check the last one first.

SEP

[David S. Bell]

We are not using async at all. This procedure worked fin under Oracle 920. This started happening with the upgrade to 10g. Is there something new in The oracle parameters that has a default of async? Do I need to add async even if I don't use it?

Dave

[Josef Watzko]

Hi,
we are seeing the same behaviour on HP-UX 11.31 with Oracle 10.2.0.4.0 too (message "OS async I/O limit ..." and slow replay of archive logs).
Someone else has a hint.

Josef

[Aneesh Mohan]

Hi,

Could you please post
#kmtune -q max_async_ports


Aneesh

[Josef Watzko]

Hi,

#kctune -q max_async_ports
Tunable Value Expression Changes
max_async_ports 4096 Default Immed

Josef