On Thu, 30 Dec 2021 at 17:18, Guillaume Lelarge <guillaume@lelarge.info> wrote: > Hello, > > I've been reading the autovacuum code (the launcher and the worker) on the > 14 branch. As previously, I've seen some configuration at the beginning, > especially for statement_timeout, lock_timeout and > idle_in_transaction_session_timeout, and I was surprised to discover there > was no configuration for idle_session_timeout. I'm not sure the code should > set it to 0 as well (otherwise I'd have written a patch), but, if there was > a decision made to ignore its value, I'd be interested to know the reason. > I could guess for the autovacuum worker (it seems to work in a transaction, > so it's already handled by the idle_in_transaction_timeout), but I have no > idea for the autovacuum launcher. > > If it was just missed, I could write a patch this week to fix this. >
Oh, it was just missed. I didn't note set autovacuum code set those settings, I think we should also set idle_session_timeout to 0.
Should we also change this for pg_dump and pg_backup_archiver?
pg_dump works in a single transaction, so it's already dealt with idle_in_transaction_timeout. Though I guess setting both would work too.