On 05 Oct 2015, at 18:04, Bruce Momjian <bruce@momjian.us> wrote:
On Mon, Oct 5, 2015 at 05:41:07PM +0200, Oleksii Kliukin wrote:
Hello,
I'm trying to find out how to rewind a cluster that was not shut down cleanly, in order to implement pg_rewind support in patroni (an automated failover system, https://github.com/zalando/patroni).
At the moment, pg_rewind limits itself to only cleanly shut down clusters. This works nicely in the case of a split brain caused by the network partitioning. However, it doesn't cover the important case of a suddenly crashed master: the crashed cluster cannot be rewound to the new master.
Did you read this thread convering the same topic from a few weeks ago?
Thanks, I saw it. The problem being discussed there is different from mine: I need to rewind a crashed master, not a replica being shut down in recovery. And I’m looking for something that be achieved with 9.3 or 9.4, and there are evidences (at least suggestions in the thread linked from my previous post) that it should work.
