Horton

RE: BUG: Former primary node might stuck when started as a standby

Hayato Kuroda (Fujitsu) <kuroda.hayato@fujitsu.com>

View thread View on postgr.es
From: "Hayato Kuroda (Fujitsu)" <kuroda.hayato@fujitsu.com>
To: 'Michael Paquier' <michael@paquier.xyz>, Alexander Lakhin <exclusion@gmail.com>
Cc: pgsql-hackers <pgsql-hackers@postgresql.org>, Aleksander Alekseev <aleksander@timescale.com>
Date: 2026-03-03T04:02:53Z
Lists: pgsql-hackers

Attachments

Dear Alexander, Michael,

> On Mon, Mar 02, 2026 at 09:00:00AM +0200, Alexander Lakhin wrote:
> > As it turned out, v2 patch works as expected, but the test may still fail
> > when the build is configured without injection points. Maybe this test (
> > and similar one(s)) should be skipped in this case, not sure...
> 
> Exactly, I don't see what else we can do here except skip the
> sequences of the test that we know may fail if injection points are
> not enabled in the build.

I had a concern that some BF animals have not enable the injection point yet
thus coverage might be decreased for them. But it's OK for me to fix it.

> Looking at v2, the patch ought to comment the reason *why* these tests
> are skipped and *why* an injection point is used.  The reader should
> have more details than just a small hint about a set of "random
> failures".  I'd suggest to document that once with the first injection
> point attached, and have the other blocks hold comments telling to
> look at the first block.

I preferred to add descriptions at the place checking enable_injection_points.
See the updated version.

Best regards,
Hayato Kuroda
FUJITSU LIMITED

Commits

  1. Fix rare instability in recovery TAP test 004_timeline_switch

  2. Fix rare instability in recovery TAP test 009_twophase