Redo Log improvement - how to analyze and attack?
Date: Wed, 14 Nov 2012 15:49:18 -0600
Message-ID: <F05D8DF1FB25F44085DB74CB916678E885650C48D9_at_NADCWPMSGCMS10.hca.corpad.net>
I have this 3-Node RAC (10.2.0.4) that I've inherited (as I've mentioned) and after putting out a few issues I'm ready to move onto the redo logs (and perhaps I should have started here earlier). First question:
Are high numbers of log file switches (during certain hours) enough validation for that an improvement is needed?
In my case, during heavy batch processing, this RAC cluster may experience 300 log switches during an hour (I know, I know - sounds bad doesn't it)
(Caveat: The log file switches were generated from within Toad - I'm assuming they are valid)
My first step was to enable FAST_START_MTTR_TARGET and it is set for 3 minutes (180 seconds).
Here's some of my numbers for the time interval on Node1:
Redo Size:
1,451,121.63 per second
AWR redo/log stats (formatted) here:
https://gist.github.com/4075050
Second question:
V$INSTANCE_RECOVERY OPTIMAL_LOGFILE_SIZE = 5421 (~5 GB)
Is OPTIMAL_LOGFILE_SIZE calculated by the *EXISTING LOGFILE GROUPS*? So if I had more groups, would this number go down?
Third Question:
I've seen various mentions of " archive_lag_target" - what do I need to think about regarding this parameter in relation to check points?
Any suggestions/thoughts are appreciated.
Regards,
Chris Taylor
Oracle DBA
Parallon IT&S
christopher.taylor2_at_parallon.net<mailto:christopher.taylor2_at_parallon.net>
www.parallon.net
-- http://www.freelists.org/webpage/oracle-lReceived on Wed Nov 14 2012 - 22:49:18 CET