Horton

Re: Minimal logical decoding on standbys

Ashutosh Sharma <ashu.coek88@gmail.com>

View thread
From: Ashutosh Sharma <ashu.coek88@gmail.com>
To: Andres Freund <andres@anarazel.de>
Cc: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com>, Robert Haas <robertmhaas@gmail.com>, Thomas Munro <thomas.munro@gmail.com>, Alvaro Herrera <alvherre@2ndquadrant.com>, Ibrar Ahmed <ibrar.ahmad@gmail.com>, Amit Khandekar <amitdkhan.pg@gmail.com>, fabriziomello@gmail.com, tushar <tushar.ahuja@enterprisedb.com>, Rahila Syed <rahila.syed@2ndquadrant.com>, pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2023-02-15T12:32:11Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Reduce the log level in 035_standby_logical_decoding.pl.

  2. 035_standby_logical_decoding: Add missing waits for replication

  3. For cascading replication, wake physical and logical walsenders separately

  4. Handle logical slot conflicts on standby

  5. Support invalidating replication slots due to horizon and wal_level

  6. Prevent use of invalidated logical slot in CreateDecodingContext()

  7. Replace replication slot's invalidated_at LSN with an enum

  8. Pass down table relation into more index relation functions

  9. Assert only valid flag bits are passed to visibilitymap_set()

  10. Remove unused _bt_delitems_delete() argument.

  11. Add xl_btree_delete optimization.

On Thu, Jan 12, 2023 at 11:46 PM Andres Freund <andres@anarazel.de> wrote:
>
> Hi,
>
> On 2023-01-12 20:08:55 +0530, Ashutosh Sharma wrote:
> > I previously participated in the discussion on "Synchronizing the
> > logical replication slots from Primary to Standby" and one of the
> > purposes of that project was to synchronize logical slots from primary
> > to standby so that if failover occurs, it will not affect the logical
> > subscribers of the old primary much. Can someone help me understand
> > how we are going to solve this problem with this patch? Are we going
> > to encourage users to do LR from standby instead of primary to get rid
> > of such problems during failover?
>
> It only provides a building block towards that. The "Synchronizing the logical
> replication slots from Primary to Standby" project IMO needs all of the
> infrastructure in this patch. With the patch, a logical rep solution can
> e.g. maintain one slot on the primary and one on the standby, and occasionally
> forward the slot on the standby to the position of the slot on the primary. In
> case of a failover it can just start consuming changes from the former
> standby, all the necessary changes are guaranteed to be present.
>
>
> > Also, one small observation:
> >
> > I just played around with the latest (v38) patch a bit and found that
> > when a new logical subscriber of standby is created, it actually
> > creates two logical replication slots for it on the standby server.
> > May I know the reason for creating an extra replication slot other
> > than the one created by create subscription command? See below:
>
> That's unrelated to this patch. There's no changes to the "higher level"
> logical replication code dealing with pubs and subs, it's all on the "logical
> decoding" level.
>
> I think this because logical rep wants to be able to concurrently perform
> ongoing replication, and synchronize tables added to the replication set. The
> pg_16399_sync_16392_7187728548042694423 slot should vanish after the initial
> synchronization.
>

Thanks Andres. I have one more query (both for you and Bertrand). I
don't know if this has already been answered somewhere in this mail
thread, if yes, please let me know the mail that answers this query.

Will there be a problem if we mandate the use of physical replication
slots and hot_standby_feedback to support minimum LD on standby. I
know people can do a physical replication setup without a replication
slot or even with hot_standby_feedback turned off, but are we going to
have any issue if we ask them to use a physical replication slot and
turn on hot_standby_feedback for LD on standby. This will reduce the
code changes required to do conflict handling for logical slots on
standby which is being done by v50-0001 and v50-0002* patches
currently.

IMHO even in normal scenarios i.e. when we are not doing LD on
standby, we should mandate the use of a physical replication slot.

--
With Regards,
Ashutosh Sharma.