Re: Performance monitoring (was: [PATCHES] Logging checkpoints and other slowdown causes)

On Sat, 12 May 2007, Jim C. Nasby wrote:

> Not to beat a dead horse, but do we really want to force folks to be
> parsing logs for performance monitoring?

All of the really interesting DBA level information about checkpoints is
now sitting in pg_stat_bgwriter. There really is no reason I'd expect
this information to be part of normal performance monitoring setup; for me
it has fit into troubleshooting and R&D.

When I'm talking about parsing logs in this context, it's for answering
questions like "how does fsync time correlate with amount of data written
at checkpoint time?" Since I've found it educational to sort through
that, I just didn't want the patch modified so that was harder to do than
it had to be.

If I thought it worth the trouble to expose that level of information via
the stats interface, I'd have submitted that instead of this log-based
patch. This circles back to the previous discussion of whether this
particular information is strictly developer-level.

--
* Greg Smith gsmith(at)gregsmith(dot)com http://www.gregsmith.com Baltimore, MD