Recovery from WAL archives not advancing timeline?

From: Don Seiler <don(at)seiler(dot)us>
To: Pgsql-admin <pgsql-admin(at)lists(dot)postgresql(dot)org>
Subject: Recovery from WAL archives not advancing timeline?
Date: 2020-08-08 06:37:06
Message-ID: CAHJZqBBk1cgbvXevUuAGTsbMTYx6_BDwwUqV_4ne_QUG=AuMmw@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-admin

PostgreSQL 9.6

I shut down the primary and promoted the replica (timeline advanced from 1
to 2).

I then put a recovery.conf on the old primary so it would recover by WAL
restore (not streaming). However it seems to refuse to see the new
timeline. I do have recovery_target_timeline='latest', my full
recovery.conf looks like this:

standby_mode = 'on'
restore_command = 'envdir /etc/wal-g/environment /usr/local/bin/wal-g
wal-fetch %f %p'
recovery_target_timeline = 'latest'

However the new replica complains:
ERROR: 2020/08/08 06:28:34.876197 openpgp: invalid argument: no armored
data found
ERROR: 2020/08/08 06:28:39.621300 Archive '000000010000142E0000005B' does
not exist.

The new primary wrote that file but under timeline 2, I
see 000000020000142E0000005B in the archive, along with the
expected 00000002.history file.

So why doesn't the new replica see that there are no more TL1 WAL files and
advance to TL2? There are still some files in its pg_xlog dir on that
replica, would it be safe to shut it down and clear them out?

--
Don Seiler
www.seiler.us

Responses

Browse pgsql-admin by date

  From Date Subject
Next Message Hotmail 2020-08-08 09:59:20 PostgreSQL JDBC Driver ERROR: invalid byte sequence for encoding "UTF8": 0x00 on Select stmt
Previous Message Tom Lane 2020-08-07 15:49:54 Re: PostgreSQL-11 partition creation issue