Re: Progress reporting for pg_verify_checksums

On Tue, Apr 02, 2019 at 08:11:45AM +0200, Fabien COELHO wrote:
> Nope, this was way before I intervened. ISTM that a patch was submitted to
> get per second or slower progress reporting on the initalization, and it was
> rejected. Now that there are many SSD, maybe I could bring it back. An issue
> with pbbench is that the reported time and progress is for the insertion
> phase only, but PK and other FK declaration take as much time and are not
> included, so I'm not sure it can be much better.

Bonus idea: integrate automatically with pg_stat_progress_vacuum() for
the last vacuum at the end of initialization. This step can also take
a very long time and one can only know the progress of the VACUUM with
a second session/terminal.

> For pg_checksums, probably some improvement patch will be submitted later,
> if someone feels like it.

Let's see. I think that what we have now in v12 is good enough for
checksum operations on an offline cluster. And my take is that we
should focus more on online checksum verification for v13 and future
versions.

Regarding all this tooling around checksums. With v12, enabling
checksums with no actual downtime is doable with a primary-standby
deployment using physical replication and one planned failover
(possible as well with v10 and newer versions and logical replication
but that takes longer), so I think that there is not much value in
being able to enable checksums on a single node while it is online.
--
Michael