Keaton Adams <kadams@mxlogic.com> writes:
> Well, no, I'm not that paranoid where I expect checkpoints to be taking
> place so often. I do find it interesting that controlling checkpoint
> frequency is a factor of the number of WAL files available. In order to
> get up to 60 second checkpoints I had to set checkpoint_segments to 25,
> which resulted in the creation of 52 log files in data/pg_xlog. So for
> 120 second checkpoints I will need to have up to 104 log files in
> data/pg_xlog available?
No, it's nonlinear, precisely because of the
checkpoints-increase-the-log-size effect. Given the size of modern
disks, I don't see why a couple Gb of WAL space is a problem anyhow
for a large database.
regards, tom lane
