On 25 August 2014 02:57, Alvaro Herrera <alvhe...@2ndquadrant.com> wrote:
> Thomas Munro wrote:
>> The difficulty of course will be testing all these racy cases reproducibly...
> Does this help?
> http://www.postgresql.org/message-id/51fb4305.3070...@2ndquadrant.com
> The useful trick there is forcing a query to get its snapshot and then
> go to sleep before actually doing anything, by way of an advisory lock.

Yes it does, thanks Alvaro and Craig.  I think the attached spec
reproduces the problem using that trick, ie shows NOWAIT blocking,
presumably in EvalPlanQualFetch (though I haven't stepped through it
with a debugger yet).  I'm afraid I'm out of Postgres hacking cycles
for a few days, but next weekend I should have a new patch that fixes
this by teaching EvalPlanQualFetch about wait policies, with isolation
tests for NOWAIT and SKIP LOCKED.

Best regards,
Thomas Munro
# Test NOWAIT with an updated tuple chain.

	id int PRIMARY KEY,
	data text NOT NULL
  INSERT INTO foo VALUES (1, 'x');


session "s1"
setup		{ BEGIN; }
step "s1a"	{ SELECT * FROM foo WHERE pg_advisory_lock(0) IS NOT NULL FOR UPDATE NOWAIT; }
step "s1b"	{ COMMIT; }

session "s2"
step "s2a"	{ SELECT pg_advisory_lock(0); }
step "s2b"	{ UPDATE foo SET data = data; }
step "s2c"	{ BEGIN; }
step "s2d"	{ UPDATE foo SET data = data; }
step "s2e"	{ SELECT pg_advisory_unlock(0); }
step "s2f"	{ COMMIT; }

permutation "s2a" "s1a" "s2b" "s2c" "s2d" "s2e" "s1b" "s2f"
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:

Reply via email to