Hi,
On 2018-09-11 12:03:44 -0400, Tom Lane wrote:
> Andres Freund <andres@anarazel.de> writes:
> > Isn't one of the most common ways to run into "out of shared memory"
> > "You might need to increase max_locks_per_transaction." to run pg_dump?
> > And that's pretty commonly done against standbys?
>
> If the startup process has acquired enough AELs to approach locktable
> full, any concurrent pg_dump has probably failed already, because it'd
> be trying to share-lock every table and so would have a huge conflict
> cross-section; it's hard to believe it wouldn't get cancelled pretty
> early in that process. (Again, if you think this scenario is probable,
> you have to explain the lack of field complaints.)
I was thinking of the other way round - there's a running pg_dump and
then somebody does a bit of DDL (say a DROP SCHEMA CASCADE in a
multi-tenant scenario).
Greetings,
Andres Freund