Re: pgstat include expansion

Amit Kapila <amit.kapila16@gmail.com>

From: Amit Kapila <amit.kapila16@gmail.com>
To: Andres Freund <andres@anarazel.de>
Cc: Alvaro Herrera <alvherre@kurilemu.de>, Chao Li <li.evan.chao@gmail.com>, pgsql-hackers@postgresql.org, Michael Paquier <michael@paquier.xyz>
Date: 2026-02-17T04:45:07Z
Lists: pgsql-hackers
On Mon, Feb 16, 2026 at 11:20 PM Andres Freund <andres@anarazel.de> wrote:
>
> I think we should simply not pass the worker type. The only reason the worker
> type passed is that that is used as a proxy for what kind of error occurred:
>
>
> /*
>  * Report a subscription error.
>  */
> void
> pgstat_report_subscription_error(Oid subid, int wtype)
> {
>     PgStat_EntryRef *entry_ref;
>     PgStat_BackendSubEntry *pending;
>
>     entry_ref = pgstat_prep_pending_entry(PGSTAT_KIND_SUBSCRIPTION,
>                                           InvalidOid, subid, NULL);
>     pending = entry_ref->pending;
>
>     switch ((LogicalRepWorkerType) wtype)
>     {
>         case WORKERTYPE_APPLY:
>             pending->apply_error_count++;
>             break;
>
>         case WORKERTYPE_SEQUENCESYNC:
>             pending->sync_seq_error_count++;
>             break;
>
>         case WORKERTYPE_TABLESYNC:
>             pending->sync_table_error_count++;
>             break;
>
>         default:
>             /* Should never happen. */
>             Assert(0);
>             break;
>     }
> }
>
>
> It doesn't seem like the right thing to have pgstat_subscription.c translate
> the worker type to the concrete error this way. Afaict each of the callsites
> of pgstat_report_subscription_error() actually knows what kind of error its
> reporting, but just uses the worker type to do so.
>
> I'd either change the signature to have one argument for each of the error
> types, i.e.
>     pgstat_report_subscription_error(int subid,
>                                      bool apply_error,
>                                      bool sequencesync_error,
>                                      bool_tablesync_error);
>
> or split the function into three, and have
>     pgstat_report_subscription_{apply,sequence,tablesync}(int subid);
>

Good idea. +1 for the second approach to split the function. We can
name them as pgstat_report_subscription_apply_error(int subid),
pgstat_report_subscription_sequence_error(int subid),
pgstat_report_subscription_tablesync_error(int subid).

With Regards,
Amit Kapila.



Commits

  1. Don't include wait_event.h in pgstat.h

  2. Don't include proc.h in shm_mq.h

  3. Don't include latch.h in libpq/libpq.h

  4. Reduce includes in pgstat.h

  5. Avoid including utils/timestamp.h in conflict.h.

  6. Avoid including worker_internal.h in pgstat.h.

  7. Add seq_sync_error_count to subscription statistics.

  8. Don't include execnodes.h in replication/conflict.h

  9. Collect statistics about conflicts in logical replication.