Обсуждение: Very long startup after crash with no message in log

Hello again,

last night we came across another unpleasant behaviour. I managed to
crash the cluster and the next start took really long (about 4 hours).
Even in single mode with -d1 there was no message telling us what's
going on. We used lsof and found out that the process opens each
database and performs some operation which takes up to a few seconds
(with almost 6k databases). It would be really nice to tell the user
what's going on. Or is there a configuration option which will enable
such logging, perhaps?

Thank you,

Ondřej

OndYej Svtlík <osvetlik@flexibee.eu> writes:

> Hello again,
>
> last night we came across another unpleasant behaviour. I managed to
> crash the cluster and the next start took really long (about 4
> hours). Even in single mode with -d1 there was no message telling us
> what's going on. We used lsof and found out that the process opens
> each database and performs some operation which takes up to a few
> seconds (with almost 6k databases). It would be really nice to tell
> the user what's going on. Or is there a configuration option which
> will enable such logging, perhaps?

Have a look at log_min_messages.

> Thank you,
>
> Ondřej

--
Jerry Sievers
Postgres DBA/Development Consulting
e: postgres.consulting@comcast.net
p: 312.241.7800