log_timestamp and SIGHUP?

Started by Andrew Sullivanabout 23 years ago5 messagesgeneral
Jump to latest
#1Andrew Sullivan
andrew@libertyrms.info

Hi,

Postgres 7.2.x.

I just set log_timestamp=false and sent the postmaster a SIGHUP. But
still I have the timestamps. Doesn't that work? The docs don't say
that this is one of the "startup only" properties.

A
--
----
Andrew Sullivan 204-4141 Yonge Street
Liberty RMS Toronto, Ontario Canada
<andrew@libertyrms.info> M2P 2A8
+1 416 646 3304 x110

#2Tom Lane
tgl@sss.pgh.pa.us
In reply to: Andrew Sullivan (#1)
Re: log_timestamp and SIGHUP?

Andrew Sullivan <andrew@libertyrms.info> writes:

I just set log_timestamp=false and sent the postmaster a SIGHUP. But
still I have the timestamps. Doesn't that work?

It sure looks like it should, modulo the fact that backends won't
re-read postgresql.conf until they next reach the
wait-for-client-command outer loop. You sure you SIGHUP'd the right
process?

regards, tom lane

#3Andrew Sullivan
andrew@libertyrms.info
In reply to: Tom Lane (#2)
Re: log_timestamp and SIGHUP?

On Tue, Mar 18, 2003 at 11:47:23PM -0500, Tom Lane wrote:

It sure looks like it should, modulo the fact that backends won't
re-read postgresql.conf until they next reach the
wait-for-client-command outer loop.

Hmm. The outer loop should surely happen after a COMMIT/ROLLBACK,
right?

You sure you SIGHUP'd the right process?

Well, assuming the postmaster.pid file wasn't lying. But yes,
because some of the other changes I made did take effect.

A

-- 
----
Andrew Sullivan                         204-4141 Yonge Street
Liberty RMS                           Toronto, Ontario Canada
<andrew@libertyrms.info>                              M2P 2A8
                                         +1 416 646 3304 x110
#4Tom Lane
tgl@sss.pgh.pa.us
In reply to: Andrew Sullivan (#3)
Re: log_timestamp and SIGHUP?

Andrew Sullivan <andrew@libertyrms.info> writes:

On Tue, Mar 18, 2003 at 11:47:23PM -0500, Tom Lane wrote:

It sure looks like it should, modulo the fact that backends won't
re-read postgresql.conf until they next reach the
wait-for-client-command outer loop.

Hmm. The outer loop should surely happen after a COMMIT/ROLLBACK,
right?

Yeah. [tries it] It works for me ...

I have another theory: if you change
#log_timestamp = false
to
log_timestamp = true
you cannot reverse that by returning the line to its original state:
the original state is a comment and won't cause the postmaster to change
its variable. You must say
log_timestamp = false
if you want to get the value changed back via SIGHUP.

(Sorry if that seems insulting to your intelligence ... but I've been
bit by it at least once myself ...)

regards, tom lane

#5Andrew Sullivan
andrew@libertyrms.info
In reply to: Tom Lane (#4)
Re: log_timestamp and SIGHUP?

On Thu, Mar 20, 2003 at 10:26:58AM -0500, Tom Lane wrote:

Yeah. [tries it] It works for me ...

Well, I'm stumped, then. It definitely did not work (on multiple
tries) on the system where I did it.

#log_timestamp = false
to
log_timestamp = true
you cannot reverse that by returning the line to its original state:

No, I know this, too (one of the folks here made exactly that error
just the other day, actually).

It's not a big deal, really. I was just curious. And annoyed that
something is up, but I don't know what.

Thanks.

A

-- 
----
Andrew Sullivan                         204-4141 Yonge Street
Liberty RMS                           Toronto, Ontario Canada
<andrew@libertyrms.info>                              M2P 2A8
                                         +1 416 646 3304 x110