Folks,
I spoke with Paul Ramsey and he suggested pgpool as a way of addressing
our apparent #httpd * #tracinstances postmaster process count.
http://pgpool.projects.postgresql.org/
I took a look and it seems promising, but I got a bit bogged down.
I've been trying to watch the number of postmaster processes. When things
are in bad shape we have on the order of 200 postmaster's running. However,
I am now seeing that when things are well behaved the number of postmasters's
fluctuates up and down and is sometimes down as low as 50 so my theory that
connections are never dropped (and thus postmasters shutdown) appears to
be wrong. I'm not sure what this implies.
Anyone interested in trying out pgpool?
I still haven't been able to file a peer1 ticket to stop running dsmc.
Best regards,
--
---------------------------------------+--------------------------------------
I set the clouds in motion - turn up | Frank Warmerdam, warmerdam@pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush | Geospatial Programmer for Rent