In a normal oracle environment you can set the FAST_START_MTTR_TARGET parameter which controls the amount of time (seconds) required to recover from an instance or server failure. 10g has a new feature Automatic Checkpoint tuning which utilizes periods of low I/O to advance checkpoints. This feature requires that FAST_START_MTTR_TARGET is set to a non-zero value.
FAST_START_MTTR_TARGET includes both instance startup and crash recovery. If you have a RAC environment and there is a failure the surviving instances will perform the recovery. Therefore including the instance startup time in the mean time to recover calculation isn't accurate. So in the case of RAC environments you may choose to set the _FAST_START_INSTANCE_RECOVERY_TARGET parameter which determines the amount of the time for instance recovery. Note: If multiple instances fail, recovery may take longer than the value you specified for this parameter.
So in summary, FAST_START_MTTR_TARGET includes the time for startup, mount, crash recovery, and open. _FAST_START_INSTANCE_RECOVERY_TARGET includes the time for instance recovery only.
Changing any parameter which affects checkpointing should be throughly tested to make sure you don't thrash your IO subsystem. Each of these parameters will increase I/O.
2 comments:
Excellent posts I enjoyed them very much. I went a step further and made them interactive by building a few 11gR2 RAC simulators myself at http://vgrigorian.com.
Thanks.
Vladimir Grigorian.
Post a Comment