Re: snapshot too old issues, first around wraparound and then more.

On Mon, Apr 13, 2020 at 5:14 PM Andres Freund <andres(at)anarazel(dot)de> wrote:
> FWIW, I think the part that is currently harder to fix is the time->xmin
> mapping and some related pieces. Second comes the test
> infrastructure. Compared to those, adding additional checks for old
> snapshots wouldn't be too hard - although I'd argue that the approach of
> sprinkling these tests everywhere isn't that scalable...

Just trying out some ideas here... I suppose the wrapping problem
just requires something along the lines of the attached, but now I'm
wondering how to write decent tests for it. Using the
pg_clobber_current_snapshot_timestamp() function I mentioned in
Robert's time->xmin thread, it's easy to build up a time map without
resorting to sleeping etc, with something like:

select pg_clobber_current_snapshot_timestamp('3000-01-01 00:00:00Z');
select pg_current_xact_id();
select pg_clobber_current_snapshot_timestamp('3000-01-01 00:01:00Z');
select pg_current_xact_id();
select pg_clobber_current_snapshot_timestamp('3000-01-01 00:02:00Z');
select pg_current_xact_id();
select pg_clobber_current_snapshot_timestamp('3000-01-01 00:03:00Z');
select pg_current_xact_id();
select pg_clobber_current_snapshot_timestamp('3000-01-01 00:04:00Z');

Then of course frozenXID can be advanced with eg update pg_database
set datallowconn = 't' where datname = 'template0', then vacuumdb
--freeze --all, and checked before and after with Robert's
pg_old_snapshot_time_mapping() SRF to see that it's truncated. But
it's not really the level of stuff we'd ideally mess with in
pg_regress tests and I don't see any precent, so I guess maybe I'll
need to go and figure out how to write some perl.