Re: postgres replication without pg_basebackup? postgres 13.3

On Mon, 2022-11-07 at 23:11 +0100, Pilar de Teodoro wrote:
> Thank you very much for the idea. [of running pg_rewind]
> We ran pg_rewind correctly:
> [postgres(at)gacsdb05 data-13.3]$ pg_rewind -c -R --target-pgdata=/PostgresDB/sas_hdd/data-13.3/  --source-server="host=XXXX port=XXXX user=postgres password=XXX" 
> pg_rewind: source and target cluster are on the same timeline
> pg_rewind: no rewind required

I cannot verify that you ran it correctly.

"Target" should be the old server with the extra transactions.
"Source" should be the promoted standby server. Promotion switches to a new timeline,
so it looks like you did something wrong.

> but then we got the following error:
>
> [2022-11-07 22:57:55 CET-]LOG:  starting PostgreSQL 13.3 on x86_64-pc-linux-gnu, compiled by gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-10), 64-bit
> [2022-11-07 22:57:55 CET-]LOG:  listening on IPv4 address "0.0.0.0", port XXXX
> [2022-11-07 22:57:55 CET-]LOG:  could not create IPv6 socket for address "::": Address family not supported by protocol
> [2022-11-07 22:57:55 CET-]LOG:  listening on Unix socket "/tmp/.s.PGSQL.XXXX"
> [2022-11-07 22:57:55 CET-]LOG:  database system was shut down in recovery at 2022-11-07 22:57:47 CET
> [2022-11-07 22:57:55 CET-]LOG:  restored log file "00000003.history" from archive
> cp: cannot stat '/PostgresWalLogArchive/new/00000004.history': No such file or directory
> [2022-11-07 22:57:55 CET-]LOG:  entering standby mode
> [2022-11-07 22:57:55 CET-]LOG:  restored log file "00000003.history" from archive
> [2022-11-07 22:57:55 CET-]LOG:  invalid primary checkpoint record
> [2022-11-07 22:57:55 CET-]PANIC:  could not locate a valid checkpoint record
> [2022-11-07 22:57:55 CET-]LOG:  startup process (PID 3011860) was terminated by signal 6: Aborted
> [2022-11-07 22:57:55 CET-]LOG:  aborting startup due to startup process failure
> [2022-11-07 22:57:55 CET-]LOG:  database system is shut down
>
> We have read we can run pg_resetwal but the Wal folder is the folder where the Wals are archived in the primary. Would that be correct to reset them?

Don't run "pg_resetwal". At the very least, it will break your standby.

At this point, your standby seems to be broken. I don't know what exactly you
did, but it leeks like you should run a "pg_basebackup" after all.

Yours,
Laurenz Albe