LGWR will not wait for checkpoint to complete?
Imad Jundi, July 23, 2017 - 5:26 am UTC
Thank you, Chris
So this means that at logswitch a full checkpoint is triggered but the lgwr will not wait for it to complete?
July 24, 2017 - 1:26 am UTC
Checkpoints have changed dramatically from the days where a log file switch was the driver for them (and hence you could get a "stalling" issue when log files switched). Nowadays you can think of checkpointing as more of a continuous activity that is always going on in the background.
But the underlying premise is the same - we cannot re-use a log file until we know that the modified blocks in the database cache are safely on disk (otherwise re-using the redo log will compromise our ability to protect those blocks).
Got it.
Imad Jundi, July 24, 2017 - 5:15 am UTC
Thank you again, Chris.
You made it very clear for me.