Re: Question about postgres replication construction

Hi Vijay,

Thanks for the reply.
The new standby does not have access to the other two nodes. I'm not aware
of being able to register it to the cluster.

By the way I've found my error. I forgot to change the archive_command on
node 2. After changing that it worked like a charm.
However i'm interested if anyone has other (better) ideas in how to tackle
this.

thanks in advance.
regards
Marian

Op vr 29 okt. 2021 om 12:06 schreef Vijaykumar Jain <
vijaykumarjain(dot)github(at)gmail(dot)com>:

> On Fri, 29 Oct 2021 at 14:07, MARIANE K <mskiesewetter(at)gmail(dot)com> wrote:
>
>> Hi everyone,
>>
>>
>>
>> I need some advice.
>>
>> I need to set up a 3 node postgres 12 cluster.
>>
>> The primary postgres server with a streaming replication hot standby
>> which is managed with repmgr.
>>
>> Besides these 2 servers I need to create another standby server which
>> will be synced with wal files that are copied to the server.
>>
>> This third node does not have any connection with the other two nodes and
>> will only be updated through the wal files.
>>
>> The third node will be created by taking a base backup of the primary
>> node and restoring it to the PGDATA directory.
>>
>> repmgr is not configured for node 3 because it has no connection with the
>> other two nodes.
>>
>> The primary connection parameter is uncommented in the postgresql.conf.
>>
>>
>>
>> This construction works until I perform a switchover from node 1 to node
>> 2 with repmgr, making node 2 the new primary and node 1 the new standby.
>>
>> The standby on node 3 does not pick up the switch.
>>
>> My question is: Does anyone have an idea in how to approach this? Is this
>> a valid construction?
>>
>> I know I'm missing something essential but at this moment I can't seem to
>> see it.
>>
>>
>>
>>
>>
> C.2. repmgr
> <https://repmgr.org/docs/current/faq-repmgr.html#FAQ-REPMGR-CLONE-OTHER-SOURCE>
>
> <https://repmgr.org/docs/current/faq-repmgr.html#FAQ-REPMGR-CLONE-OTHER-SOURCE>you
> would want to register the new standby to the cluster.
>
>
> --
> Thanks,
> Vijay
> Mumbai, India
>