Re: [PATCHES] Fix "database is ready" race condition

On Feb 5, 2007, at 8:19 AM, Tom Lane wrote:
> "Simon Riggs" <simon(at)2ndquadrant(dot)com> writes:
>> My suggestions would be
>> 1. "Database system has completed recovery" and
>> 2. "Database system is ready to accept connections"
>
> The second was in fact the wording I had in mind, sorry for not being
> clear. As to the first, the question is whether a log message at that
> specific point has any use. It's not "completion of recovery",
> exactly,
> since we go through that spot whether it's a normal startup or
> recovery
> (and there is another log message when we actually do any WAL replay).
> AFAICS it's a fairly randomly chosen place in a long sequence of
> Things
> That Must Happen. Part of the reason Markus is seeing a race
> condition
> is that this isn't the last thing done before the startup subprocess
> exits --- see BootstrapMain. So I'm for just getting rid of it.

It is useful to know if the database had to do recovery, though, and
if it did do recovery, it would be useful to know how long it took if
the subsequent startup took a real amount of time.

BTW, this is a real problem I've seen on a database with 500k entries
in pg_class... it takes several minutes to finish starting after the
'Postmaster is ready' message.
--
Jim Nasby jim(at)nasby(dot)net
EnterpriseDB http://enterprisedb.com 512.569.9461 (cell)