On Fri, Aug 11, 2017 at 2:20 PM, Don Seiler <don@seiler.us> wrote:
> On Fri, Aug 11, 2017 at 11:07 AM, Scott Marlowe <scott.marlowe@gmail.com>
> wrote:
>>
>> If you're doing minor upgrades, you don't need to bring down the whole
>> cluster, you can take a single replica out and upgrade pg on it, then
>> put it back in the cluster. Again, if you have enough kept wal
>> segments, or you use replication slots, then it will just catch right
>> up.
>
>
> What do you mean by "take it out of the cluster" and "put it back in the
> cluster"? Shut it down and rename the recovery.conf? Or more complexity than
> that? Why would that be needed at all?
Yeah, basically touch the the file defined to make it fail over so it
becomes its own master.
Then test upgrade it and test the crap out of it separate from
production. If all works fine, then resubscribe it and now you know
your upgrades in production should work.