Обсуждение: archive_timeout behavior (8.4.6)

Hi all,
     We thought we were having issues with our warm standby (postgres
8.4.6), but think we figured it out, however I'd like someone to confirm
what we think is true.

If a cluster has zero activity whatsoever, but we have archive_timeout
set (say to 60 seconds), will it create a WAL file anyways, or knowing
there was zero activity does it skip it? We turned it on (temporarily)
to force WAL files to go to a newly set up warm standby so we could
verify that we have the whole end-to-end process correct, but we didn't
have any WAL files transfer to the warm standby. After some trouble
shooting, we notice that we do get a WAL file shortly after we insert a
single row into a table.

The confusion is probably because I remember this same scenario on older
versions of 8.4, and watching it generate wal files even with no activity.

So long story short, if a cluster has zero activity, and archive_timeout
is not set to 0, when it reaches that timeout should it generate a file,
or just skip because it would be 100% empty?


Thanks in advance,
     Brian F

On 04/22/2011 08:23 PM, Brian Fehrle wrote:
> So long story short, if a cluster has zero activity, and
> archive_timeout is not set to 0, when it reaches that timeout should
> it generate a file, or just skip because it would be 100% empty?

You're right about what you've seen here, there is some logic in the
WAL/checkpoint code that can avoid doing anything when the timeout hits,
if there hasn't been any new WAL written since the last one.  That
hasn't changed for a while now though (since I've been watching that
part of the source code at least) so I'm not sure what differences
between 8.4.6 and earlier 8.4 versions you think you've observed.

--
Greg Smith   2ndQuadrant US    greg@2ndQuadrant.com   Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support  www.2ndQuadrant.us
"PostgreSQL 9.0 High Performance": http://www.2ndQuadrant.com/books