Horton

Re: [bug fix] Cascading standby cannot catch up and get stuck emitting the same message repeatedly

Tsunakawa, Takayuki <tsunakawa.takay@jp.fujitsu.com>

View thread View on postgr.es
From: "Tsunakawa, Takayuki" <tsunakawa.takay@jp.fujitsu.com>
To: 'Michael Paquier' <michael.paquier@gmail.com>
Cc: "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>
Date: 2016-08-26T02:57:52Z
Lists: pgsql-hackers 
From: pgsql-hackers-owner@postgresql.org
[mailto:pgsql-hackers-owner@postgresql.org] On Behalf Of Michael Paquier
> 9.3 has addressed that by allowing streaming standbys to perform timeline
> jumps via the replication protocol. Doesn't this problem enter in this area?

IIUC, that new feature enables timeline switch without disconnecting the standby and without WAL archive.  I think 9.2 can perform timeline switch with WAL archive.  In fact, the customer said they hit the problem only once in many occurrences of the same test.  The bug seems to emerge depending on the timing.

Regards
Takayuki Tsunakawa

Commits

  1. Fix bug in verifying TLI (timeline ID) in WAL page header during recovery..

  2. Split out XLog reading as an independent facility