Re: [pgeu-general] Replication failover - Mailing list pgsql-general

From TJ
Subject Re: [pgeu-general] Replication failover
Date
Msg-id 519DCB70.9020105@melodicninja.co.uk
Whole thread Raw
Responses Re: [pgeu-general] Replication failover  (Heikki Linnakangas <hlinnakangas@vmware.com>)
List pgsql-general
We have a few different sets of servers with different versions.
9.0.4
9.1.4
9.2.3

I recently tried to  fail-over a set of 9.2.3 servers and server4 did
notice the timeline change but did not start following it.
We do not have the recovery_target_timeline set in the recovery.conf

ATM we are not using WAL archiving only streaming.

And server4 is behind server3.

So would we need to setup WAL archiving to achieve this?

- TJ

On 22/05/13 23:22, Heikki Linnakangas wrote:
> On 22.05.2013 10:23, TJ wrote:
>> I am looking to migrate my databases from one set of hardware to another
>> all the servers are running PGSQL 9.x
>
> Which version exactly? There are differences in this area between 9.0,
> 9.1, and 9.2. If you can choose, use the latest 9.2.X version.
>
>> The problem i have is that when i fail over to server3 using the trigger
>> file it will increment the timeline which will stop the replication of
>> server4.
>
> With 9.1 and above, it will notice the new timeline and start
> following it, if you set up a WAL archive in addition to streaming
> replication and set recovery_target_timeline='latest' in
> recovery.conf. Starting with 9.3, a standby can follow a timeline
> changes over streaming replication too, so in 9.3 it should just work.
>
> That's all assuming that server4 is behind server3; if it has already
> replayed WAL beyond the point in the WAL where server3 switched to a
> new timeline, it can't follow that timeline switch.
>
> - Heikki
>
>



pgsql-general by date:

Previous
From: Sergey Konoplev
Date:
Subject: Re: Rule for all the tables in a schema
Next
From: Heikki Linnakangas
Date:
Subject: Re: [pgeu-general] Replication failover