Re: fixing failed master after standby promotion

On Mon, Aug 8, 2016 at 11:35 AM, Masahiko Sawada <sawada(dot)mshk(at)gmail(dot)com> wrote:
> On Sun, Aug 7, 2016 at 10:55 PM, Aviel Buskila <aviel33(at)gmail(dot)com> wrote:
>> Hey ,
>> I have been setting a configuration of 2 nodes (master and standby)
>> replicating using repmgr, and 1 server using for load balancing using
>> pgpool-II.
>> Now after the master has failed, the standby was successfully promoted.
>> I would like to know if there is any elegant way to semi-automatically
>> resync the failed master and re-register him as a standby server?
>>
>
> I'm not familiar with repmgr but you can use the pg_rewind in order to
> automatically synchronize a data cluster if you use PostgreSQL 9.5 or
> later.
> In 9.4 or before, you need to take whole backup from the new master
> server, for example using pg_basebackup.

Note that this needs wal_log_hints to be set to on so as pages having
their hint bints modified get WAL-logged after a checkpoint so as
pg_rewind is able to track such pages. pg_rewind has been integrated
if Postgres core in 9.5, but note that I spend some time maintaining a
version of pg_rewind for 9.3 and 9.4 [1], and it is proving to be in a
rather stable state. I got that integrated into some of the things I
work on, and nobody complains about it.

[1]: https://github.com/vmware/pg_rewind
--
Michael