Re: New VACUUM FULL

Jeff Davis <pgsql(at)j-davis(dot)com> writes:
> I'd like to try to summarize the goals we have for VACUUM to put these
> patches in perspective:

Good summary, but ...

> Some other ideas also came out of the thread, like:

> * Provide a way to truncate the dead space from the end of a relation in
> a blocking manner. Right now, lazy vacuum won't shrink the file unless
> it can acquire an exclusive lock without waiting, and there's no way to
> actually make it wait. This can be a separate command, not necessarily a
> part of VACUUM.

What I remembered was actually closer to the opposite: people are
concerned about lazy vac holding the exclusive lock too long once it
does acquire it. In a manual vacuum this leads to unexpected blockage
of concurrent queries, and in an autovac this leads to a forced cancel
preventing autovac from completing the operation (which means no
space removal at all, and no stats update either). The code is designed
on the assumption that it won't spend very long holding the ex lock,
and so I think a reasonable TODO item is to ensure that by having it
limit how many blocks it will scan during the shrinkage attempt.

FWIW, once we provide the extensible option syntax, it doesn't seem
like it'd be hard to have an option to make it wait for the lock,
so I'd recommend that approach over anything with a separate command.

regards, tom lane