On 4 May 2017 at 20:45, Magnus Hagander <magnus@hagander.net> wrote:
> On Thu, May 4, 2017 at 2:42 PM, Craig Ringer <craig@2ndquadrant.com> wrote:
>>
>> On 4 May 2017 at 20:05, Magnus Hagander <magnus@hagander.net> wrote:
>> > PFA a patch that adds a new function, pg_move_replication_slot, that
>> > makes
>> > it possible to move the location of a replication slot without actually
>> > consuming all the WAL on it.
>>
>> > This can be useful for example to keep replication slots in sync between
>> > different servers in a replication cluster.
>>
>> It needs a test to ensure it only operates on physical slots. It
>> should ERROR on a logical slot, since it has no way of correctly
>> advancing the catalog_xmin or finding a reasonable restart_lsn for
>> logical decoding.
>
>
> I guess that makes sense, yeah. I didn't try it with that :)
Barring that and matching docs changes, looks fine to me at first glance.
Not sure you need to acquire the spinlock on the slot, since you
acquired the slot for your backend. It won't hurt, but I don't think
it's necessary.
I'm not convinced you need a WARNING for moving the slot backwards. If
one is emitted, it should be a proper ereport with current position
and requested position in errdetail. I'm not sure it's a useful
message though.
-- Craig Ringer http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services