Re: [COMMITTERS] pgsql: Allow a streaming replication standby to follow a timeline switc - Mailing list pgsql-general

From Heikki Linnakangas
Subject Re: [COMMITTERS] pgsql: Allow a streaming replication standby to follow a timeline switc
Date
Msg-id 50D30984.2010204@vmware.com
Whole thread Raw
In response to Re: [COMMITTERS] pgsql: Allow a streaming replication standby to follow a timeline switc  (hubert depesz lubaczewski <depesz@depesz.com>)
Responses Re: [COMMITTERS] pgsql: Allow a streaming replication standby to follow a timeline switc
List pgsql-general
On 18.12.2012 13:42, hubert depesz lubaczewski wrote:
> In pg_log on ubuntu2 I see:
>
> 2012-12-18 12:41:34.428 CET [unknown]@[unknown] 1685  LOG:  connection received: host=172.28.173.142 port=45842
> 2012-12-18 12:41:34.430 CET replication@[unknown] 1685 172.28.173.142(45842) LOG:  replication connection authorized:
user=replication
> 2012-12-18 12:41:34.432 CET replication@[unknown] 1685 172.28.173.142(45842) ERROR:  requested WAL segment
000000020000000000000015has already been removed 
> 2012-12-18 12:41:34.433 CET replication@[unknown] 1685 172.28.173.142(45842) LOG:  disconnection: session time:
0:00:00.005user=replication database= host=172.28.173.142 port=45842 
>
> Something looks weird. To put it lightly.

Hmm, that's a different error than you got before. Thom also reported a
"requested WAL segment ... has already been removed" error, but in his
test case, and as far as I could reproduce it, the error doesn't reoccur
when the standby reconnects. In other words, it eventually worked
despite that error. In any case, I just committed a fix for the scenario
that Thom reported. Can you try again with a fresh checkout?

- Heikki


pgsql-general by date:

Previous
From: Sergey Konoplev
Date:
Subject: Re: DONT_CARE Aggregate
Next
From: Robert James
Date:
Subject: Re: DONT_CARE Aggregate