Thread: pgpool on Opteron server running FreeBSD 6.0
I'm running postgres 8.1.3 on a 2-CPU dual-core Opteron server with 8-Gigs of RAM. The server's operating system is FreeBSD 6.0. We migrated our client applications from Informix and went through all of the standard steps to resolve differences between the two databases packages. However, there are a handful of applications that generate connection storms which we hoped could be mitigated by using pgpool. We installed pgpool (num_init_children = 32 and max_pool = 1)ran an application requiring only 14 pgpool children and observed the following: a) The system load is > 20 b) WCPU values for the 14 active pgpool processes exceed 20% c) Context switching on the server jumps as high as 250,000 The application now takes 4 (with pgpool) rather than less than two hours (without pgpool) to complete (it took less than an hour with Informix). Strange! Any ideas? Thanks -- -------------------------------------------------------------------------------- Andrew Rost National Operational Hydrologic Remote Sensing Center (NOHRSC) National Weather Service, NOAA 1735 Lake Dr. West, Chanhassen, MN 55317-8582 Voice: (952)361-6610 x 234 Fax: (952)361-6634 andy.rost@noaa.gov http://www.nohrsc.noaa.gov --------------------------------------------------------------------------------
On Sep 7, 2006, at 3:31 PM, andy rost wrote: > a) The system load is > 20 > b) WCPU values for the 14 active pgpool processes exceed 20% > c) Context switching on the server jumps as high as 250,000 > The application now takes 4 (with pgpool) rather than less than two > hours (without pgpool) to complete (it took less than an hour with > Informix). Strange! Any ideas? we're going production with a similar setup (though using 64-bit Xeon) running FreeBSD 6.1/amd64 and pgpool. I'll check for any funny behavior like this, but in pre-production testing we are not seeing any interrupt storms or hight context switch.