Home » Server Options » Data Guard » tunning the MRP process
tunning the MRP process [message #294173] Wed, 16 January 2008 11:59 Go to next message
aline
Messages: 92
Registered: February 2002
Member
Hi,
I have some problem with my physical standby database.
The MRP process is working slowly.
The primary redo generation is greater than the redo apply rate.
I'm working with a 9.2 database.

do you know how to tune this process or if there is some documentations about that?

thk!
Re: tunning the MRP process [message #306714 is a reply to message #294173] Sun, 16 March 2008 02:37 Go to previous messageGo to next message
kumarkis
Messages: 5
Registered: March 2008
Junior Member
we had the same problem...we changed a few parameters and it worked well...can u send me (at kishore131@gmail.com) the parameter files of both the servers,server configuration(no. of cpus,memory) and logsize,generation rate...
i hope we can tune it
Re: tunning the MRP process [message #306717 is a reply to message #306714] Sun, 16 March 2008 03:12 Go to previous messageGo to next message
Michel Cadot
Messages: 68624
Registered: March 2007
Location: Nanterre, France, http://...
Senior Member
Account Moderator
Why don't you explain what you did to improve the performances?
Not just the parameters you modified but why.

Regards
Michel
icon14.gif  Re: tunning the MRP process [message #306729 is a reply to message #306717] Sun, 16 March 2008 05:58 Go to previous messageGo to next message
kumarkis
Messages: 5
Registered: March 2008
Junior Member
Hi,

I wud like to share my experiences:

As per oracle's best recovery practices,
we had changed the keep pool,recycle pool to zero and instead allocated that to db_cache_size.

If the recovery is done by parallel processes, intercommunication b/w processes can be improved by increasing the parallel_execution_message_size. Its new value depends on requirement. We had increased it to 32KB to good benefit.

Pls check if these are applicable in your case.

Regards,
skk
Re: tuning the MRP process [message #308901 is a reply to message #306729] Tue, 25 March 2008 22:00 Go to previous message
RichB
Messages: 3
Registered: March 2008
Location: California
Junior Member
Hi,

We use LNS to ship the archivelogs asynchronously.

We configured the LISTENER.ORA to increase the Send and receive buffers on both the primary and standby. There's a formula to calculate the optimum based on your RTT time of a ping average.
(SEND_BUF_SIZE=10500000)
(RECV_BUF_SIZE=10500000)

We also run multiple archive processes log_archive_max_processes=13 ... 1 for the local archive and 12 to ship to the standby. The 12 are used in the LOG_ARCHIVE_DEST_2 with a MAX_CONNECTIONS=4. This will parallelize the shipping of each archlog. so essentially, we will ship 3 logs at a time ,each broken up into 4 streams).

And set the ARCHIVE_LAG_TARGET = 600 to perform a log switch every 600 seconds regardless of activity... for our SLA of a maximum of 10 minutes of data loss in a true destructive disaster.

BTW, we're on 10gR2.

We're shipping several hundred GB a day this way from California to Texas on multiple databases.

Hope this helps.
Rich

Previous Topic: How many standy database should I build?
Next Topic: ora 16055
Goto Forum:
  


Current Time: Thu Mar 28 06:29:56 CDT 2024