Re: Application of archived logs on physical standby very slow
Date: Fri, 29 Jan 2010 11:04:01 +0000
Hi Luca et al,
thanks for all your input.
In my case it doesn't seem to be Oracle causing the problem, it's the MSA 1000. I don't know much about its configuration but it seems (to say it in layman's terms) that the storage admins added more disks and grouped them into a disk group. They then carved out 2 logical volumes and presented one of these to the DR box and the other one to production. Now production is terribly busy (the system grows 2-3 GB/day!), and after a reorganisation essentially all IO ends up on that newly added LUN. So this affects all the disks in the disk group on the MSA which in term means slow IO response times for DR.
Golden rule of not sharing SAN/disks etc violated once more.
-- > Martin Bach > OCM 10g > http://martincarstenbach.wordpress.com > http://www.linkedin.com/in/martincarstenbach Luca Canali wrote:Received on Fri Jan 29 2010 - 05:04:01 CST
> What do you see in V$RECOVERY_PROGRESS as "Active Apply Rate"? (see also discussion in http://jarneil.wordpress.com/2008/07/15/monitoring-your-standby-with-vrecovery_progress/)
> I have 35655 KB/s on one of my systems for example.
> I had a look at the discussion on redo apply tuning in Larry Carpenter's (et al.) book at page 119 [one can find it on google books too in case paper in not available :)] they mention several diagnostic and tuning techniques, but on the subject of MRP scan/read of logs they say "This part of the recovery is easily handled by the single MRP process".
> PS: in terms of further investigations http://www.oracle.com/technology/deploy/availability/pdf/MAA_WP_10gRecoveryBestPractices.pdf seems to suggest setting Log_Archive_Trace=8192 for some 'advance troubleshooting'