Horton

Re: Subscription statistics are not dropped at DROP SUBSCRIPTION in some cases

Nathan Bossart <nathandbossart@gmail.com>

View thread View on postgr.es
From: Nathan Bossart <nathandbossart@gmail.com>
To: Masahiko Sawada <sawada.mshk@gmail.com>
Cc: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2023-05-09T00:07:21Z
Lists: pgsql-hackers 
On Mon, May 08, 2023 at 04:23:15PM +0900, Masahiko Sawada wrote:
> We call pgstat_drop_subscription() at the end of DropSubscription()
> but we could leave from this function earlier e.g. when no slot is
> associated with the subscription. In this case, the statistics entry
> for the subscription remains. To fix it, I think we need to call it
> earlier, just after removing the catalog tuple. There is a chance the
> transaction dropping the subscription fails due to network error etc
> but we don't need to worry about it as reporting the subscription drop
> is transactional.

Looks reasonable to me.  IIUC calling pgstat_drop_subscription() earlier
makes no real difference (besides avoiding this bug) because it is uѕing
pgstat_drop_transactional() behind the scenes.

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

Commits

  1. Stabilize subscription stats test.

  2. pgstat: fix subscription stats entry leak.

  3. Create subscription stats entry at CREATE SUBSCRIPTION time

  4. pgstat: drop subscription stats without slot as well, fix comment