On Sun, Jun 2, 2019 at 11:47 AM Adrian Klaver <adrian.klaver@xxxxxxxxxxx> wrote:
On 6/1/19 8:07 PM, Tom K wrote:
>
> https://www.postgresql.org/docs/10/app-postgres.html
> Single-User Mode
> ...
>
> and see if that at least gets the server started. This is a highly
> restricted so do not expect much usability.
>
>
> These servers did crash before however didn't' notice anything
> particularly wrong at the time. Perhaps a better check was in order.
>
> Here is the result from attempting to start the DB in single user mode.
> -bash-4.2$ /usr/pgsql-10/bin/postgres --single -D /data/patroni
> --config-file=/data/patroni/postgresql.conf
> 2019-06-01 23:06:28.679 EDT [31969] LOG: database system was
> interrupted; last known up at 2019-04-28 06:06:24 EDT
> 2019-06-01 23:06:28.953 EDT [31969] LOG: invalid record length at
> 0/4C35CDF8: wanted 24, got 0
> 2019-06-01 23:06:28.953 EDT [31969] LOG: invalid primary checkpoint record
> 2019-06-01 23:06:28.953 EDT [31969] LOG: using previous checkpoint
> record at 0/4C34EDA8
> 2019-06-01 23:06:28.953 EDT [31969] PANIC: replication checkpoint has
> wrong magic 0 instead of 307747550
That comes from origin.c in the logical replication section of the code.
Did you have logical replication set up?
Nope. wal_level was set to replica, not logical. Unless you mean
What was the role of this cluster in the original setup?
The cluster was the backend database for a number of applications. The aim was to point applications to a single large cluster instead of a number of small instances of postgres running all over the LAB.
If I can get through the above error and get to the DB's and tables, I could just dump them out and reinit the entire DB from the backup.
> Aborted
> -bash-4.2$
>
--
Adrian Klaver
adrian.klaver@xxxxxxxxxxx
