Isn't remote_write a really dumb name for that setting?
AFAICT, the remote_write setting for synchronous_commit is named exactly
backwards, because the point of the setting is that it *doesn't* wait
for the remote to write anything.
As an alternative I suggest "remote_receive". Perhaps somebody else
has a better idea?
regards, tom lane
On Wed, Aug 22, 2012 at 01:01:04PM -0400, Tom Lane wrote:
AFAICT, the remote_write setting for synchronous_commit is named exactly
backwards, because the point of the setting is that it *doesn't* wait
for the remote to write anything.As an alternative I suggest "remote_receive". Perhaps somebody else
has a better idea?
Yes, I didn't like remote_write either; see this thread:
http://archives.postgresql.org/pgsql-hackers/2012-05/msg00375.php
Yes, please, I would like it changed.
--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://enterprisedb.com
+ It's impossible for everything to be true. +
Bruce Momjian <bruce@momjian.us> writes:
On Wed, Aug 22, 2012 at 01:01:04PM -0400, Tom Lane wrote:
AFAICT, the remote_write setting for synchronous_commit is named exactly
backwards, because the point of the setting is that it *doesn't* wait
for the remote to write anything.As an alternative I suggest "remote_receive". Perhaps somebody else
has a better idea?
Yes, I didn't like remote_write either; see this thread:
http://archives.postgresql.org/pgsql-hackers/2012-05/msg00375.php
Oh. After re-reading that thread (and looking at the code to be sure),
I think the mode name is all right; it's the documentation that is 100%
broken. The actual meaning of the setting is that we wait for the
remote to write() the data, but not fsync() it. The description in the
SGML docs has nothing to do with reality.
Will fix the docs.
regards, tom lane