On Mon, 2026-07-13 at 17:04 +0530, Dilip Kumar wrote:
> PFA a patch for skipping subscriptions belonging to different
> databases.
I suggest we add a few checks (or asserts) that subdbid==MyDatabaseId
when reading pg_subscription from anywhere other than the launcher.
Also, the REASSIGN OWNED path goes through
AlterSubscriptionOwner_oid(), whereas both the REASSIGN OWNED and ALTER
SUBSCRIPTION ... OWNER TO paths go through
AlterSubscriptionOwner_internal(). It would be more clear to add the
skipping logic in the former and an assert in the latter.
Other than that, LGTM.
We can also reconsider how we handle dependencies to address Noah's
concern, but this patch is good regardless.
Regards,
Jeff Davis
