On Tue, Aug 19, 2014 at 8:22 PM, Andres Freund <andres@2ndquadrant.com> wrote:
> On 2014-08-19 19:59:51 +0900, Fujii Masao wrote:
>> On Tue, Aug 19, 2014 at 7:21 PM, Andres Freund <andres@2ndquadrant.com> wrote:
>> > On 2014-08-19 08:21:10 +0800, Craig Ringer wrote:
>> >> On 08/19/2014 01:03 AM, Robert Haas wrote:
>> >> > 2. I agree that it's not good to have this get controlled by a GUC.
>> >> > If the behavior change is big enough that it's going to break clients,
>> >> > adding a GUC isn't a sufficient remedy. If it's not, adding a GUC is
>> >> > unnecessary.
>> >>
>> >> There's plenty of agreement on "not a GUC" - but what about alternatives?
>> >
>> > What's the problem with the COMMIT WITH (report_lsn on) I've proposed?
>> > Reporting the LSN in the command tag? Anything doing transparent
>> > failover needs to be aware of transaction boundaries anyway.
>>
>> So something like transparent failover doesn't work when a client is
>> working in auto commit mode? That sounds not good.
>
> I don't think transparent failover + autocommit is a sensible
> combination.
>
>> Just idea. What about using NoticeResponse message to report LSN?
>> It can be sent basically anytime and this idea doesn't break current
>> wire protocol.
>
> I think that'd be horrible from multiple perspectives: a) how to discern
> them from regular notice messages
You can implement your own protocol upon existing messages like
replication is done.
b) It's not sent in the same protocol
> level message as the COMMIT message. Thus there's scenarios where you
> only have the commit, but not the LSN.
Hmm.. you can change the code so that the message with LSN is sent
as soon as COMMIT message is sent, if required.
Regards,
--
Fujii Masao
