Horton

Re: wake up logical workers after ALTER SUBSCRIPTION

Nathan Bossart <nathandbossart@gmail.com>

View thread View on postgr.es
From: Nathan Bossart <nathandbossart@gmail.com>
To: Amit Kapila <amit.kapila16@gmail.com>
Cc: Melih Mutlu <m.melihmutlu@gmail.com>, Thomas Munro <thomas.munro@gmail.com>, "Hayato Kuroda (Fujitsu)" <kuroda.hayato@fujitsu.com>, "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>
Date: 2023-01-05T04:12:37Z
Lists: pgsql-hackers 
On Thu, Jan 05, 2023 at 09:09:12AM +0530, Amit Kapila wrote:
> On Wed, Jan 4, 2023 at 11:03 PM Nathan Bossart <nathandbossart@gmail.com> wrote:
>> On Wed, Jan 04, 2023 at 09:41:47AM +0530, Amit Kapila wrote:
>> > If so, we probably also need to
>> > ensure that table_states_valid is marked false probably via
>> > invalidations so that we can get the latest state and then perform
>> > this check. I guess if we can do that then we can directly move the
>> > restart logic to the end.
>>
>> IMO this shows the advantage of just waking up the worker.  It doesn't
>> change the apply worker's behavior besides making it more responsive.
> 
> But there doesn't appear to be any guarantee that the result for
> AllTablesyncsReady() will change between the time it is invoked
> earlier in the function and at the place you have it in the patch.
> This is because the value of 'table_states_valid' may not have
> changed. So, how is this supposed to work?

The call to CommandCounterIncrement() should set table_states_valid to
false if needed.

-- 
Nathan Bossart
Amazon Web Services: https://aws.amazon.com

Commits

  1. Avoid type cheats for invalid dsa_handles and dshash_table_handles.

  2. Track logrep apply workers' last start times to avoid useless waits.

  3. Check for two_phase change at end of process_syncing_tables_for_apply.

  4. Wake up a subscription's replication worker processes after DDL.