From: avinashvallarapu (@avinashvallarapu) <noreply+avinashvallarapu@github.com>
To: pgpartman/pg_partman <noreply+pgpartman-pg_partman@github.com>
Subject: Re: [pgpartman/pg_partman] issue #860: Feature: p_skip_on_error parameter for run_maintenance_proc()
Date: Sun, 04 Oct 2026 21:52:14 +0000
Message-ID: <gh-pgpartman-pg_partman-860-c5984787481@github.com> (raw)
In-Reply-To: <gh-pgpartman-pg_partman-860@github.com>
References: <gh-pgpartman-pg_partman-860@github.com>
Looks like 5.5.0 already covers this. `run_maintenance()` now catches errors per partition set, logs a WARNING, sets `maintenance_last_run` to NULL for that set and moves on (the CVE-2026-61822 change). `run_maintenance_proc()` calls it once per set, so one broken set no longer stops the others.
I tried it on the `551` branch with PG 18.6: two sets, the first with a bad `partition_interval` and `maintenance_order = 1`. After `CALL partman.run_maintenance_proc()` the first was skipped with the warning and the second went from 6 to 11 children.
Side note on the code in the issue: PL/pgSQL doesn't allow `COMMIT` or `ROLLBACK` inside a block that has an `EXCEPTION` clause, so the commit has to stay outside it, the way the procedure does it now.
I think this one can be closed.