Horton

Re: promotion related handling in pg_sync_replication_slots()

shveta malik <shveta.malik@gmail.com>

View thread View on postgr.es
From: shveta malik <shveta.malik@gmail.com>
To: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
Cc: "Zhijie Hou (Fujitsu)" <houzj.fnst@fujitsu.com>, Amit Kapila <amit.kapila16@gmail.com>, Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com>, pgsql-hackers <pgsql-hackers@postgresql.org>, shveta malik <shveta.malik@gmail.com>
Date: 2024-04-18T12:06:05Z
Lists: pgsql-hackers

Attachments

On Thu, Apr 18, 2024 at 12:35 PM shveta malik <shveta.malik@gmail.com> wrote:
>
> To fix above issues, these changes have been made in v7:

Please find v8 attached. Changes are:

1) It fixes ShutDownSlotSync() issue, where we perform
kill(SlotSyncCtx->pid). There are chances that after we release
spin-lock and before we perform kill, slot-sync worker has error-ed
out and has set SlotSyncCtx->pid to InvalidPid (-1) already. And thus
kill(-1) could result in abnormal process kills on some platforms.
Now, we get pid under spin-lock and then use it to perform kill to
avoid pid=-1 kill. This is on a similar line of how ShutdownWalRcv()
does it.

2) Improved comments in code.

3) Updated commit message with new fixes. I had missed to update it in
the previous version.

thanks
Shveta

Commits

  1. Post-commit review fixes for slot synchronization.

  2. Allow synced slots to have their inactive_since.