On Tue, 2021-04-13 at 02:38 -0700, Ma Xinjian wrote:
> When I use pg_basebackup to backup and restore db(Let's call it A) to a
> standalone instance(Let's call it B), "missing chunk number 0 for toast
> value xxx in pg_toast_xxx" errors output.
>
> PG version: 10.3
> pg_basebackup command:
> /usr/pgsql-10/bin/pg_basebackup -h p-rdb-c01 -D /var/lib/pgsql/10/data
> -Xs -P -n --waldir=/tmp/pg_wal
> I have mounted a disk to /tmp/pg_wal before, then I will mount the disk
> to /var/lib/pgsql/10/data/pg_wal, so as to ensure completeness of wal
> records during backup.
>
> Since I don't want B to be a standy server, I just want it to be a
> standalone server.
> I removed recovery.conf, then simply start postgresql-10.service. It turned
> out that postgresql-10.service
> can be started successfully. But when I use this postgresql(reindex, vacumm
> and so on), "missing chunk number 0 for toast value xxx in pg_toast_xxx"
> errors output.
>
> When pg_basebackup, it will store wal under pg_wal, can't postgresql work
> with wal records locally?
> I think primary_conninfo in recovery.conf is just used to get newer wal
> records from A. Right?
>
> I have also tested:
> If I start postgresql-10.service with recovery.conf firstly, then split it
> from postgresql cluster, everything works fine.
>
> Above test seems proved that it is wal records's problem. I am really
> confused.
Your mail got me confused...
Why do you write the WAL to /tmp/pg_wal, only to later mount that at the
default location?
I see nothing wrong with what you are doing, but I may have got lost
in your complicated procedure.
You don't happen to remove "backup_label", do you?
Yours,
Laurenz Albe
--
Cybertec | https://www.cybertec-postgresql.com
