Horton

Re: speed up a logical replica setup

Amit Kapila <amit.kapila16@gmail.com>

View thread
From: Amit Kapila <amit.kapila16@gmail.com>
To: "Hayato Kuroda (Fujitsu)" <kuroda.hayato@fujitsu.com>
Cc: Alexander Lakhin <exclusion@gmail.com>, Peter Eisentraut <peter@eisentraut.org>, Euler Taveira <euler@eulerto.com>, Shlok Kyal <shlok.kyal.oss@gmail.com>, Tomas Vondra <tomas.vondra@enterprisedb.com>, Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com>, "pgsql-hackers@lists.postgresql.org" <pgsql-hackers@lists.postgresql.org>, Michael Paquier <michael@paquier.xyz>, Andres Freund <andres@anarazel.de>, Ashutosh Bapat <ashutosh.bapat.oss@gmail.com>, Fabrízio de Royes Mello <fabriziomello@gmail.com>, vignesh C <vignesh21@gmail.com>
Date: 2024-07-03T06:20:40Z
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. pg_createsubscriber: Remove obsolete comment

  2. pg_createsubscriber: Fix an unpredictable recovery wait time.

  3. Fix unstable test in 040_pg_createsubscriber.

  4. Fix the testcase introduced in commit 81d20fbf7a.

  5. Further weaken new pg_createsubscriber test on Windows.

  6. Temporarily(?) weaken new pg_createsubscriber test on Windows.

  7. Make pg_createsubscriber warn if publisher has two-phase commit enabled.

  8. Make pg_createsubscriber more wary about quoting connection parameters.

  9. pg_createsubscriber: Remove failover replication slots on subscriber

  10. pg_createsubscriber: Remove replication slot check on primary

  11. pg_createsubscriber: Only --recovery-timeout controls the end of recovery process

  12. pg_createsubscriber: creates a new logical replica from a standby server

  13. Add some const decorations

  14. Add option force_initdb to PostgreSQL::Test::Cluster:init()

  15. Remove MSVC scripts

On Wed, Jul 3, 2024 at 10:42 AM Hayato Kuroda (Fujitsu)
<kuroda.hayato@fujitsu.com> wrote:
>
> Based on that, I considered a scenario why the slot could not be synchronized.
> I felt this was not caused by the pg_createsubscriber.
>
> 1. At initial stage, the xmin of the physical slot is 743, and nextXid of the
>    primary is also 743.
> 2. Autovacuum worker starts a new transaction. nextXid is incremented to 744.
> 3. Tries to creates a logical replication slot with failover=true *before the
>    transaction at step2 is replicated to the standby*.
> 4. While creating the slot, the catalog_xmin must be determined.
>    The initial candidate is nextXid (= 744), but the oldest xmin of replication
>    slots (=743) is used if it is older than nextXid. So 743 is chosen in this case.
>    This operaion is done in CreateInitDecodingContext()->GetOldestSafeDecodingContext().
> 5. After that, the transaction at step2 is reached to the standby node and it
>    updates the nextXid.
> 6. Finally runs pg pg_sync_replication_slots() on the standby. It finds a failover
>    slot on the primary and tries to create on the standby. However, the
>    catalog_xmin on the primary (743) is older than the nextXid of the standby (744)
>    so that it skips to create a slot.
>
> To avoid the issue, we can disable the autovacuuming while testing.
>

Your analysis looks correct to me. The test could fail due to
autovacuum. See the following comment in
040_standby_failover_slots_sync.

# Disable autovacuum to avoid generating xid during stats update as otherwise
# the new XID could then be replicated to standby at some random point making
# slots at primary lag behind standby during slot sync.
$publisher->append_conf('postgresql.conf', 'autovacuum = off');

> # Descriptions for attached files
>
> An attached script can be used to reproduce the first failure without pg_createsubscriber.
> It requires to modify the code like [1].

> 0003 patch disables autovacuum for node_p and node_s. I think node_p is enough, but did
> like that just in case. This fixes a second failure.
>

Disabling on the primary node should be sufficient. Let's do the
minimum required to stabilize this test.

-- 
With Regards,
Amit Kapila.