| From: | Bertrand Drouvot <bertranddrouvot(dot)pg(at)gmail(dot)com> |
|---|---|
| To: | Nathan Bossart <nathandbossart(at)gmail(dot)com> |
| Cc: | pgsql-hackers(at)postgresql(dot)org |
| Subject: | Re: pg_restore fails on extracted tar archives with large objects |
| Date: | 2026-11-20 08:04:19 |
| Message-ID: | av//A7/fudljFB52@bdtpg |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Hi,
On Thu, Nov 19, 2026 at 03:23:32PM -0600, Nathan Bossart wrote:
> The following commands produce an error:
>
> createdb src
> createdb dst
> psql src -c "select lo_from_bytea(1, '\xff')"
> pg_dump -Ft -f d.tar src
> mkdir ex
> tar xf d.tar -C ex
> pg_restore -Fd -d dst ex
>
> pg_restore: error: could not open large object TOC file "ex/2966.dat" for input: No such file or directory
>
> This is because the tar format puts NNN.dat instead of blobs_NNN.toc into
> the BLOBS TOC entry. That was harmless until commit a45c78e328 (v17)
> taught the directory format's restore code to use the file name in the TOC
> entry. Commit 166b4f4560 fixed a similar problem for directory-format
> dumps made by versions before v12.
>
> The attached patch fixes it by constructing the file name instead of using
> the one in the TOC entry.
It looks like _PrepParallelRestore() is affected by the same issue. The stat()
fails, leaving the BLOBS entry’s dataLength at zero, so the parallel scheduler
cannot use the intended size estimate. Should this use the same filename
construction as _LoadLOs()?
Regards,
--
Bertrand Drouvot
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com
| From | Date | Subject | |
|---|---|---|---|
| Previous Message | Michael Paquier | 2026-11-20 08:00:49 | Re: Checkpoints stuck after a failed bulk write inside a savepoint |