Dear experts,
we are running Oracle 10.2.0.5 on RHEL5.
It's DB with absolutely no load on it, just my tests executed. Recently checking parameters around checkpoint I decided to see the relationship between FAST_START_MTTR_TARGET and checkpoint frequency.
I have set this as parameters associated with checkpoint:
LOG_CHECKPOINTS_TO_ALERT=TRUE
FAST_START_MTTR_TARGET=5
log_checkpoint_timeout=0
I'm doing an update on a large table to generate a load. I don't query any stats regarding load / volume of changes I made, but I know that 4 runs of update command make me a switch of redo group that has a size of 200M.
I'm monitoring alert file to see the notification of possible checkpoint and also use this query to identify if dbfs are updated with new CKPT SCN:
select distinct(CHECKPOINT_CHANGE#) from v$datafile;
Problem is that in my test I always have redo log switch and associated checkpoint faster than any possible checkpoint caused by oracle to respect fast_start_mttr_target.
Can you please let me know what is wrong in my test scenario, or the way I'm checking the fact of a checkpoint.
Could it be that 1 redo log file holds too few changes, that I'm always behind this 5 seconds border?
Thanks in advance.