Re: Increasing timeout of poll_query_until for TAP tests

On Mon, Jul 25, 2016 at 2:38 PM, Alvaro Herrera
<alvherre(at)2ndquadrant(dot)com> wrote:
> Michael Paquier wrote:
> Yeah, thanks, pushed. However this doesn't explain all the failures we see:

I missed those ones, thanks for the reminder.

> 1) In
> http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=hamster&dt=2016-07-14%2016%3A00%3A06
> we see the pg_basebackup test failing. I suppose that failure is also
> because of slowness, though of course this patch won't fix it.

That's from 010_pg_basebackup... And I am not sure what's behind that.
Could it be possible to add --verbose to the commands of
pg_basebackup? We may be able to catch the problem if it shows up
again.

> 2) In
> http://buildfarm.postgresql.org/cgi-bin/show_stage_log.pl?nm=hamster&dt=2016-06-29%2016%3A00%3A06&stg=recovery-check
> we see a completely different failure:
>
> error running SQL: 'psql:<stdin>:1: ERROR: relation "tab_int" does not exist
> LINE 1: SELECT count(*) FROM tab_int
> ^'
> while running 'psql -XAtq -d port=52824 host=/tmp/or2xHglniM dbname=postgres -f - -v ON_ERROR_STOP=1' at /home/buildfarm/data/buildroot/HEAD/pgsql.build/src/test/recovery/../../../src/test/perl/PostgresNode.pm line 1166.
>
> Do we have an explanation for this one?

Ah, yes, and that's a stupid mistake. We had better use
replay_location instead of write_location. There is a risk that
records have not been replayed yet even if they have been written on
the standby, so it is possible that the query looking at tab_int may
not see this relation.
--
Michael