Re: In-order pg_dump (or in-order COPY TO)

On 8/30/25 18:21, Dimitrios Apostolou wrote:
> Sorry I was not remembering the details. Probably there is a TOC in your dump file, but it does not contain any positions for the data. The pg_restore command has to scan the whole file in advance, and fill in the TOC offsets in memory.
>
> This scanning happens in a very inefficient way, with many seek calls and small block reads. Try strace to see them. This initial phase can take hours in a huge dump file, before even starting any actual restoration.

It may be that my coffee is not strong enough, but I don't understand
what you are trying to say.

Are you using, from previous post, the following?:

"Two patches for speeding up scanning an archive without TOC, like the
one I'm having (because it is piped into borg, instead of written to
file). These were activated, but shouldn't matter. They only build the
TOC in pg_restore's memory. "

The part I don't see is how you get a dump file without a TOC?

When I do the pg_dump and pipe it to Borg the resulting file has a TOC.

Can you show the rest of the | borg create ... command?

>
> Thank you for testing.
> Dimitris
>
> On 30 August 2025 20:19:13 CEST, Adrian Klaver <adrian(dot)klaver(at)aklaver(dot)com> wrote:

--
Adrian Klaver
adrian(dot)klaver(at)aklaver(dot)com