Re: Too many WAL(s) despite low transaction

Selva,

* Selva manickaraja (mavles78(at)gmail(dot)com) wrote:
> If our check_timeout is 30 minutes, what would be an acceptable time limit
> for archive_timeout?

They're two different things. Checkpoints are about getting data
flushed out to the data files (so they're not just in the WALs),
archive_timeout is about how often WAL segments should be forcibly
archived (so that the archive server doesn't end up missing data on
low-write systems).

Typically, I'd pick archive_timeout of around 5m or 10m, depending on
how much time you don't mind losing. I'd also compress the WALs (on a
low-write system, they're going to have very little data in them).
There's also a utility out there, iirc, which will truncate WALs to
remove empty space.

> Also since bulk loading/migration of large amount of data was done earlier,
> do I need to run vacuum etc.

Erm, you should be running autovacuum..

Stephen