Home > mailing lists

Re: pg_dump crashes - Mailing list pgsql-general

From	Nico De Ranter
Subject	Re: pg_dump crashes
Date	June 2, 2020 14:23:32
Msg-id	CALVv0fbqWgwxNPx7MURXQDc_pDYGd8furV7jGpRF_t9NJ+qJ9A@mail.gmail.com Whole thread Raw
In response to	Re: pg_dump crashes (Adrian Klaver <adrian.klaver@aklaver.com>)
List	pgsql-general

Tree view

FYI: I tried setting the md5 field to '' in the whole table but that didn't fix the pg_dump issue. In the end I decided to drop the database and revert to my last successful backup. I'm now still reading in all tapes to reconstruct the latest database state.

Thanks for the help anyway.

Nico

On Mon, May 25, 2020 at 4:20 PM Adrian Klaver <adrian.klaver@aklaver.com> wrote:

On 5/24/20 10:30 PM, Nico De Ranter wrote:
> Unfortunately not. I discovered the issue rather late. The last working
> backup is about 2 months old.

Well first it is entirely possible this is not the only corruption in
the database.

Second you are probably going to have to reach out to the Bacula folks:

https://www.bacula.org/support/
https://sourceforge.net/projects/bacula/lists/bacula-users

I would say the questions to ask them are:

1) Is the md5 required for a file?

2) If so how and over what is calculated?

Then you could experiment with trying to update the md5 field with new
data.

> >
> >
> >
> > Following up on the max(bigint), I tried
> >
> > SELECT md5 FROM public.file where fileid >2087994666;
> >
> > and got
> >
> > ERROR: compressed data is corrupted
> >
> > So it does look like those entries are killing it. Now for the
> > million-dollar question: how do I get them out?
>
> Do you have recent previous backup?
>
> >
> > Nico
> >
> > --
> >
> > Nico De Ranter
> >
> > Operations Engineer
> >
>
>
>
> --
> Adrian Klaver
> adrian.klaver@aklaver.com <mailto:adrian.klaver@aklaver.com>
>
>
>
> --
>
> Nico De Ranter
>
> Operations Engineer
>

--
Adrian Klaver
adrian.klaver@aklaver.com