You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by Dean Gaudet <dg...@arctic.org> on 1998/03/26 06:33:03 UTC

process spawning and such

I've had a tweak on one of my servers for a while that logs when
children are killed off due to idleness.  It's really quite noisy.
Here's a sample:

[Wed Mar 25 15:03:43 1998] [info] server has 34 children, 21 are idle, killing one
[Wed Mar 25 15:03:45 1998] [info] server has 33 children, 21 are idle, killing one
[Wed Mar 25 15:03:46 1998] [info] server has 33 children, 22 are idle, killing one
[Wed Mar 25 15:03:47 1998] [info] server has 32 children, 23 are idle, killing one
[Wed Mar 25 15:03:48 1998] [info] server has 32 children, 23 are idle, killing one
[Wed Mar 25 15:03:52 1998] [info] server has 31 children, 22 are idle, killing one
[Wed Mar 25 15:03:53 1998] [info] server has 31 children, 21 are idle, killing one
[Wed Mar 25 15:03:54 1998] [info] server has 30 children, 21 are idle, killing one
[Wed Mar 25 15:03:55 1998] [info] server has 30 children, 21 are idle, killing one
[Wed Mar 25 15:04:00 1998] [info] server has 29 children, 27 are idle, killing one
[Wed Mar 25 15:04:01 1998] [info] server has 29 children, 27 are idle, killing one
[Wed Mar 25 15:04:02 1998] [info] server has 28 children, 25 are idle, killing one
[Wed Mar 25 15:04:03 1998] [info] server has 28 children, 26 are idle, killing one
[Wed Mar 25 15:04:04 1998] [info] server has 27 children, 26 are idle, killing one
[Wed Mar 25 15:04:05 1998] [info] server has 27 children, 26 are idle, killing one
[Wed Mar 25 15:04:06 1998] [info] server has 26 children, 25 are idle, killing one
[Wed Mar 25 15:04:07 1998] [info] server has 26 children, 24 are idle, killing one
[Wed Mar 25 15:04:08 1998] [info] server has 25 children, 24 are idle, killing one
[Wed Mar 25 15:04:09 1998] [info] server has 25 children, 24 are idle, killing one
[Wed Mar 25 15:04:10 1998] [info] server has 24 children, 23 are idle, killing one
[Wed Mar 25 15:04:11 1998] [info] server has 24 children, 21 are idle, killing one
[Wed Mar 25 15:04:12 1998] [info] server has 23 children, 22 are idle, killing one
[Wed Mar 25 15:04:13 1998] [info] server has 23 children, 22 are idle, killing one
[Wed Mar 25 15:04:14 1998] [info] server has 22 children, 21 are idle, killing one
[Wed Mar 25 15:04:15 1998] [info] server has 22 children, 21 are idle, killing one
[Wed Mar 25 15:04:16 1998] [info] server has 21 children, 21 are idle, killing one
[Wed Mar 25 15:04:17 1998] [info] server has 21 children, 21 are idle, killing one

Now when I added that I thought it would be useful for tuning the
Min/MaxSpare.  I used to have MaxSpare set at 16... and got lots of
things like above, so I raised it to 20.  I still get this stuff above 20.

I dunno I don't think that the 1.3 process model is completely broken
but I've had a hard time convincing myself it doesn't do something
really silly.  You know for example, it wakes up every second and scans
the scoreboard, even if the server isn't doing anything.  This isn't
new in 1.3, but it does seem kind of silly on servers where apache is
just a sideshow.

If anyone else has any ideas on how to maintain the pool of preforked
children... I'd love to hear 'em.  This stuff is still important in
threaded models because I assume we'll have a pool of threads too.

Hmm, maybe my real problem is the other end -- maybe I should decrease
MinSpare from 8 to 4... I'll go try that and see how it behaves.  Maybe
it's just too sensitive about creating new children.

Dean