| From: | Jakub Wartak <jakub(dot)wartak(at)enterprisedb(dot)com> |
|---|---|
| To: | Chao Li <li(dot)evan(dot)chao(at)gmail(dot)com> |
| Cc: | PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org> |
| Subject: | Re: repack: clarify final phase of concurrent mode in file header comment |
| Date: | 2026-06-10 08:33:36 |
| Message-ID: | CAKZiRmy__XBiGEYAQzsAujAZ-HUE_25Tr2tUVFuVDbWOQwDZZw@mail.gmail.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On Thu, Apr 16, 2026 at 5:33 AM Chao Li <li(dot)evan(dot)chao(at)gmail(dot)com> wrote:
>
> Hi,
>
> The file header comment in repack.c briefly explains the workflow of concurrent mode as:
>
> 1. Lock the table with ShareUpdateExclusiveLock
> 2. Do an initial copy as non-concurrent mode
> 3. While the initial copy is going on, accumulates concurrent changes to a file
> 4. After the initial copy, read the file and re-apply the changes
> 5. Upgrade to AccessExclusiveLock and swap the relfilenodes
>
> When I read the comment, I got confused at step 4. While that re-apply pass is running, concurrent changes can still occur, so what happens to those changes?
>
> After reading the code, I see that there is actually an extra step after upgrading to AccessExclusiveLock to process the remaining changes.
>
> I am afraid other readers may have the same confusion, so I am proposing this small patch to make the file header comment a bit clearer.
LGTM, so +1
-J.
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Xuneng Zhou | 2026-06-10 08:36:14 | Re: t/035_standby_logical_decoding.pl might fail on attempt to read wrong timeline |
| Previous Message | Álvaro Herrera | 2026-06-10 08:33:35 | Re: Fix unqualified catalog references in psql describe queries |