Re: Invalid Assert while validating REPLICA IDENTITY?

On Wed, 11 Sept 2024 at 10:05, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
>
> On Tue, Sep 10, 2024 at 2:16 PM Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
> >
> > On Tue, Sep 10, 2024 at 11:36 AM vignesh C <vignesh21(at)gmail(dot)com> wrote:
> > >
> > > On Mon, 9 Sept 2024 at 13:12, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
> > > >
> > > > The second part of the assertion is incomplete. The
> > > > IsIndexUsableForReplicaIdentityFull() should be used only when the
> > > > remote relation has REPLICA_IDENTITY_FULL set. I haven't tested all
> > > > possible cases yet but I think the attached should be a better way to
> > > > write this assertion.
> > >
> > > The changes look good to me.
> > >
> >
> > Thanks, I'll push this tomorrow unless Dilip or anyone else has any
> > comments. BTW, as the current code doesn't lead to any bug or
> > assertion failure, I am planning to push this change to HEAD only, let
> > me know if you think otherwise.
> >
>
> Pushed.

I noticed that Drongo failed the subscription-check for the 015_stream
test, as reported at [1].

In publisher log I could see the following statement executed
015_stream.pl line 253:
$node_publisher->safe_psql('postgres', "TRUNCATE TABLE test_tab_2");

With the following log content from 015_stream_publisher.log:
2024-09-11 09:13:36.886 UTC [512:4] 015_stream.pl LOG: statement:
TRUNCATE TABLE test_tab_2

and lastly executed wait_for_catchup 015_stream.pl line 254:
2024-09-11 09:13:38.811 UTC [4320:4] 015_stream.pl LOG: statement:
SELECT '0/20E3BC8' <= replay_lsn AND state = 'streaming'
FROM pg_catalog.pg_stat_replication
WHERE application_name IN ('tap_sub', 'walreceiver')

In subscriber log I could see the following statement executed
015_stream.pl line 255:
$node_subscriber->safe_psql('postgres',
"CREATE UNIQUE INDEX idx_tab on test_tab_2(a)");

With the following log content from 015_stream_subscriber.log:
2024-09-11 09:13:39.610 UTC [3096:4] 015_stream.pl LOG: statement:
CREATE UNIQUE INDEX idx_tab on test_tab_2(a)

However, there is no clear indication of what transpired afterward or
why the remaining statements were not executed. It appears this issue
is unrelated to the recent patch, as the changes introduced by the
patch affect only update and delete operations.
I will keep monitoring the buildfarm for additional runs to determine
if the issue recurs and draw conclusions based on the observations.

[1] - https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=drongo&dt=2024-09-11%2007%3A24%3A53

Regards,
Vignesh