Alvaro Herrera wrote:
> Pushed to all affected branches, along with a somewhat lame
> isolationtester test for the condition (since we've already broken this
> twice and not noticed for long).
Buildfarm member okapi just failed this test in 9.4:
2018-01-04 02:44:43.481 PST [5a4e059a.3af1:6] ERROR: deadlock detected
2018-01-04 02:44:43.481 PST [5a4e059a.3af1:7] DETAIL: Process 15089 waits for
ShareLock on virtual transaction 4/882; blocked by process 15090.
Process 15090 waits for ShareLock on virtual transaction 3/906; blocked
by process 15089.
Process 15089:
CREATE INDEX CONCURRENTLY mcic_one_pkey ON mcic_one (id)
WHERE lck_shr(281457);
Process 15090:
CREATE INDEX CONCURRENTLY mcic_two_pkey ON mcic_two (id)
WHERE unlck();
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=okapi&dt=2018-01-04%2010%3A35%3A02
The interesting thing here is that the failure seems to be just like
before the patch: each session is waiting on a snapshot from the other
session. I suppose this must be some *other* snapshot, not the catalog
snapshot ...
--
Álvaro Herrera https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
