On 26 Mar 2015, at 19:48, Scott Ribe <scott_ribe@elevated-dev.com> wrote:
>
>> On Mar 26, 2015, at 12:42 PM, Scott Ribe <scott_ribe@elevated-dev.com> wrote:
>>
>> On Mar 26, 2015, at 12:17 PM, Graeme B. Bell <grb@skogoglandskap.no> wrote:
>>>
>>> ...I won't be able to directly apply the (promoted) standby's new WAL entries over the top of it.
>>
>> I see--there's our difference. When I do this, I am willing to stay on the standby for a while if need be.
>>
>>> A checkpoint or autovacuum might generate a small change/entry in WAL (I don't know this for sure regarding
autovacuum;this is a worst case assumption).
>>
>> I would think autovacuum would have to, since it writes some changes to at least index pages.
>
> But you can disable autovacuum. And you can manually checkpoint. So maybe you'd just add that after shutting down
servicesthat access the db. (For me, that's mostly: "sudo launchctl unload my.particular.prefix.*")
That may work, but this is starting to look very hacky, and the problem with hacky approaches is that you usually miss
somethingyou don't know about yet, or get caught by something that changes later.
For example: in the processes that you were using prior to our conversation, which you mentioned before (shutting down
servicesbefore server), were you already doing a forced checkpoint to counter archive_timeout, and had you disabled
autovacuum?
Graeme.
