Re: Fix assert failure when decoding XLOG_PARAMETER_CHANGE on primary
Masahiko Sawada <sawada.mshk@gmail.com>
From: Masahiko Sawada <sawada.mshk@gmail.com>
To: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
Cc: Amit Kapila <amit.kapila16@gmail.com>, PostgreSQL-development <pgsql-hackers@postgresql.org>
Date: 2025-02-05T18:59:22Z
Lists: pgsql-hackers
Attachments
- v2-0001-Fix-assertion-when-decoding-XLOG_PARAMETER_CHANGE.patch (application/octet-stream) patch v2-0001
On Tue, Feb 4, 2025 at 11:48 PM Bertrand Drouvot
<bertranddrouvot.pg@gmail.com> wrote:
>
> Hi,
>
> On Wed, Feb 05, 2025 at 12:08:26PM +0530, Amit Kapila wrote:
> > On Wed, Feb 5, 2025 at 2:06 AM Masahiko Sawada <sawada.mshk@gmail.com> wrote:
> > >
> > > I've attached the updated patch. The fix needs to be back-patched to
> > > v16 where logical decoding on standby was introduced.
>
> Nice catch and thanks for the patch!
> I also do prefer the current idea (disallow to start if the hot_standby is off
> and there is an existing logical slot).
>
> > Isn't it better to give the new ERROR near the below code?
> > RestoreSlotFromDisk()
> > {
> > ...
> > if (cp.slotdata.database != InvalidOid && wal_level < WAL_LEVEL_LOGICAL)
> > ereport(FATAL,
> > (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
> > errmsg("logical replication slot \"%s\" exists, but \"wal_level\" <
> > \"logical\"",
> > NameStr(cp.slotdata.name)),
> > ...
> > }
> >
> > If feasible, this will avoid an additional loop over the slots and a
> > new ERROR will be added at the same place as an existing similar
> > ERROR.
>
> Agree.
+1
>
> + if (SlotIsLogical(s) && !EnableHotStandby)
> + ereport(FATAL,
> + (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
> + errmsg("hot standby must be enabled for pre-existing logical replication slots")));
>
> On a primary lowering wal_level < logical, we'd get something like:
>
> "
> FATAL: logical replication slot "logical_slot" exists, but "wal_level" < "logical"
> "
>
> What about being close to it, so something like?
>
> "
> FATAL: logical replication slot "logical_slot" exists, but "hot_standby" = "off"
> "
Looks good, but I'd like to mention that this is required only on
standbys. So how about the following?
FATAL: logical replication slot "s" exists on the standby, but
"hot_standby" = "off"
HINT: Change "hot_standby" to be "on".
I've updated the patch accordingly.
Regards,
--
Masahiko Sawada
Amazon Web Services: https://aws.amazon.com
Commits
-
Fix assertion when decoding XLOG_PARAMETER_CHANGE on promoted primary.
- cc628f66187c 16.9 landed
- 174952ece1c0 17.5 landed
- 48796a98d5ae 18.0 landed