Re: ALTER TABLE .. DETACH PARTITION CONCURRENTLY

On 2021-May-24, Noah Misch wrote:

> prairiedog and wrasse failed a $SUBJECT test after this (commit 8aba932).
> Also, some non-CLOBBER_CACHE_ALWAYS animals failed a test before the fix.
> These IsolationCheck failures match detach-partition-concurrently[^\n]*FAILED:
>
> sysname │ snapshot │ branch │ bfurl
> ────────────┼─────────────────────┼────────┼────────────────────────────────────────────────────────────────────────────────────────────────

Checking this list, these three failures can be explained by the
detach-partition-concurrently-3 that was just patched.

> wrasse │ 2021-04-21 10:38:32 │ HEAD │ http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=wrasse&dt=2021-04-21%2010%3A38%3A32
> prairiedog │ 2021-04-25 22:05:48 │ HEAD │ http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=prairiedog&dt=2021-04-25%2022%3A05%3A48
> wrasse │ 2021-05-11 03:43:40 │ HEAD │ http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=wrasse&dt=2021-05-11%2003%3A43%3A40

Next there's a bunch whose error message is the same that we had seen
earlier in this thread; these animals are all CLOBBER_CACHE_ALWAYS:

step s1insert: insert into d4_fk values (1);
+ERROR: insert or update on table "d4_fk" violates foreign key constraint "d4_fk_a_fkey"

> hyrax │ 2021-03-27 07:29:34 │ HEAD │ http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=hyrax&dt=2021-03-27%2007%3A29%3A34
> trilobite │ 2021-03-29 18:14:24 │ HEAD │ http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=trilobite&dt=2021-03-29%2018%3A14%3A24
> hyrax │ 2021-04-01 07:21:10 │ HEAD │ http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=hyrax&dt=2021-04-01%2007%3A21%3A10
> avocet │ 2021-04-05 15:45:56 │ HEAD │ http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=avocet&dt=2021-04-05%2015%3A45%3A56
> hyrax │ 2021-04-06 07:15:16 │ HEAD │ http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=hyrax&dt=2021-04-06%2007%3A15%3A16
> hyrax │ 2021-04-11 07:25:50 │ HEAD │ http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=hyrax&dt=2021-04-11%2007%3A25%3A50
> hyrax │ 2021-04-20 18:25:37 │ HEAD │ http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=hyrax&dt=2021-04-20%2018%3A25%3A37

This is fine, because the fix commit 8aba932 is dated April 22 and these
failures all predate that.

And finally there's these two:

> topminnow │ 2021-03-28 20:37:38 │ HEAD │ http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=topminnow&dt=2021-03-28%2020%3A37%3A38
> dragonet │ 2021-04-01 19:48:03 │ HEAD │ http://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=dragonet&dt=2021-04-01%2019%3A48%3A03

(animals not CCA) which are exposing the same problem in
detach-partition-concurrently-4 that we just fixed in
detach-partition-concurrently-3, so we should apply the same fix: add a
no-op step right after the cancel to prevent the error report from
changing. I'll go do that after grabbing some coffee.

Thanks for digging into the reports!

--
Álvaro Herrera 39°49'30"S 73°17'W
"Cada quien es cada cual y baja las escaleras como quiere" (JMSerrat)