Folks,
I've been testing varying SPINS_PER_DELAY in a client's installation of
PostgreSQL against a copy of a production database, to test varying this
statistic as a way of fixing the issue.
It does not seem to work.
I've tested all of the following graduated levels:
100 (the original)
250
500
1000
2000
5000
10000
20000
30000
50000
None of these quantities seem to make any difference at all in the number of
context switches -- neither down nor up. Seems to me like this is a dead
end. Does anyone have test results that show otherwise?
--
-Josh Berkus
Aglio Database Solutions
San Francisco