| From: | surya poondla <suryapoondla4(at)gmail(dot)com> |
|---|---|
| To: | shihao zhong <zhong950419(at)gmail(dot)com> |
| Cc: | michael(at)paquier(dot)xyz, alvherre(at)kurilemu(dot)de, pgsql-hackers(at)lists(dot)postgresql(dot)org |
| Subject: | Re: Partitioned index marked valid while a partition is pending detach |
| Date: | 2026-11-12 00:04:15 |
| Message-ID: | CAOVWO5rA_OqBu40R21nHR-0r8SFBMbXfFe--UVGEV9wKyHi5dg@mail.gmail.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Hi Shihao,
Thanks for v4. I tested it on the current master, it applies cleanly, make
check and the
isolation tests pass.
It fixes the 0003 cases I sent( SET STATISTICS, ALTER COLUMN TYPE) now
reaches p1 and FINALIZE works
afterwards. It also works when p1 is itself partitioned.
I agree with leaving the 0001-only invalid index for a separate patch,
since plain DETACH has the same gap.
Two small things on 0003:
1) The spec covers ADD CHECK and SET DEFAULT, but not the cases that
actually broke before. Could you add ALTER COLUMN TYPE and DROP COLUMN on
the parent followed by FINALIZE?
2) Parent DDL now locks the pending partition too, so a session that is
only reading p1 now blocks ALTER TABLE on t, and anything queued behind it
on t. That makes sense, but I think the commit message
should mention it, and maybe a sentence in the FINALIZE part of
alter_table.sgml saying ALTER TABLE on the parent also applies to a
partition pending detach. Just a suggestion not a blocker for me.
Overall the patch looks good.
Regards,
Surya Poondla
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Zsolt Parragi | 2026-11-12 00:05:35 | Re: Interrupts vs signals |
| Previous Message | Sami Imseih | 2026-11-12 00:01:30 | Re: Improve pg_stat_statements scalability |