Tom Lane schrieb am 08.01.2018 um 16:35:
> OK. The explanation then presumably is that something is dropping that
> table concurrently with the pg_dump run. It remains unclear to me why
> pg_dump is failing to protect itself against such a drop, but before we
> can figure that out, we need to be able to reproduce the situation.
>
>
> Can you identify anything that would be dropping tables concurrently
> with this script?
I feel ashamed :(
There was a left over pipe symbol in the script that we didn't see:
pg_dump --table=l10n --table=l10n_value --clean ${db_stag} |
psql --single-transaction --dbname=${db_prod} --file=${backup_stag}
So pg_dump pipes a script with DROP statement into psql, which then
runs essentially the same script again which was created in a previous step (also with --clean)
Seeing my (stupid) mistake, I am not sure if this is supposed to work or not...
Thomas