Thread: Logfile created when not needed?
Digging into several win32 related issues right now, so I don't have the time to investigate ATM, but I figured I should get it out here: If I configure log_destination='eventlog', and then redirect_stderr='on', PostgreSQL will attempt to create a logfile in pg_log anyway. This file remains empty (at least through my test), but it is created, and if it fails to be created, the postmaster won't start at all. I may be missing something and there is a point to this, but if it's never used it should probably not be created in the first place, should it? (I would assume the same thing happens if you set log_destination='syslog', but haven't tested that) //Magnus
"Magnus Hagander" <mha@sollentuna.net> writes:
> If I configure log_destination='eventlog', and then
> redirect_stderr='on', PostgreSQL will attempt to create a logfile in
> pg_log anyway.
So don't do that --- redirect_stderr is only sensible to turn on if you
mean to use logging to stderr.
The reason we don't attempt to enforce this is that log_destination is
legal to change after startup, whereas we can't change redirect_stderr
on-the-fly (because existing backends wouldn't have their stderr tied to
the right place).
I'm not sure how sensible it really is to switch dynamically between
logging to stderr and logging someplace else, but as long as the code
can support that, why not allow it ...
regards, tom lane
> > If I configure log_destination='eventlog', and then > > redirect_stderr='on', PostgreSQL will attempt to create a > logfile in > > pg_log anyway. > > So don't do that --- redirect_stderr is only sensible to turn > on if you mean to use logging to stderr. I don't. Normally. But I had it turned on, then changed log_destination and forgot.. > The reason we don't attempt to enforce this is that > log_destination is legal to change after startup, whereas we > can't change redirect_stderr on-the-fly (because existing > backends wouldn't have their stderr tied to the right place). Ok. Fair 'nuf. //Magnus