Re: Proposal: "Causal reads" mode for load balancing reads without stale data

On Wed, Mar 30, 2016 at 6:04 AM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> On Tue, Mar 29, 2016 at 3:17 AM, Michael Paquier
> <michael(dot)paquier(at)gmail(dot)com> wrote:
>> OK, so I am switching this patch as "Ready for committer", for 0001.
>> It is in better shape now.
>
> Well... I have a few questions yet.
>
> The new argument to SyncRepWaitForLSN is called "bool commit", but
> RecordTransactionAbortPrepared passes true. Either it should be
> passing false, or the parameter is misnamed or at the least in need of
> a better comment.
>
> I don't understand why this patch is touching the abort paths at all.
> XactLogAbortRecord sets XACT_COMPLETION_SYNC_APPLY_FEEDBACK, and
> xact_redo_abort honors it. But surely it makes no sense to wait for
> an abort to become visible.

You're right, that was totally unnecessary. Here is a version that
removes that (ie XactLogAbortRecord doesn't request apply feedback
from the standby, xact_redo_abort doesn't send apply feedback to the
primary and RecordTransactionAbortPrepared now passes false to
SyncRepWaitForLSN so it doesn't wait for apply feedback from the
standby). Also I fixed a silly bug in SyncRepWaitForLSN when capping
the mode. I have also renamed XACT_COMPLETION_SYNC_APPLY_FEEDBACK to
the more general XACT_COMPLETION_APPLY_FEEDBACK, because the later
0004 patch will use it for a more general purpose than
synchronous_commit.

--
Thomas Munro
http://www.enterprisedb.com