04.12.2018 5:19, Michael Paquier wrote:
On Tue, Dec 04, 2018 at 01:28:15PM +1300, Thomas Munro wrote:
On Tue, Dec 4, 2018 at 1:18 PM Michael Paquier <mich...@paquier.xyz> wrote:
Yes, you are right. It should be "on" as "remote_flush" is not a valid
value. remote_flush is listed in SyncCommitLevel though, so this makes
me wonder if we should also introduce a new value for this purpose
available for users. The fix you propose looks good to me. Any
opinions from others?
+1 for the patch.
Thanks for confirming, Thomas. I'll go apply hopefully tomorrow if
nobody has objections.
As for introducing remote_flush as the true name of the level, this
was discussed but somehow went off-course and never made it to the
finish line:
https://www.postgresql.org/message-id/flat/CAEepm%3D3FFaanSS4sugG%2BApzq2tCVjEYCO2wOQBod2d7GWb%3DDvA%40mail.gmail.com
Oh, I forgot this one. We may want to revive that... remote_flush is
more meaningful than on, especially since there are more and more
possible values for synchronous_commit.
Yeah, I think the notion *remote_flush level* is more appropriate
especially in the context of sync replication. Within this context maybe
it makes sense to replace the word *level* to *value* in description of
*flush_lag*?
--
Regards,
Maksim Milyutin
