tracking trac ... or something

1 view
Skip to first unread message

Matt Read

unread,
Nov 18, 2009, 4:52:22 PM11/18/09
to habar...@googlegroups.com
Trac has been very flaky lately, which I believe was caused by too few
fcgi process allowed to spawn for https traffic. The problem was when
we increased that, the resource consumption got crazy. So trac is now
running on https only, and all http traffic is redirected to https.

The reason for this is that in order to run http AND https you
basically end up running two instances of trac at the same time; that
means we need to spawn enough fcgi processes to handle the load. This
made apache consume about 830 MB of ram just to keep enough procs
around to be somewhat stable. If we were to assume 1/2 the people
would use http, and the other 1/2 use https, we could cut each in
half, but I doubt that would be the case.

So now there is only one trac running, with 500 request per process,
and at minimum 4 processes ready to go (pls don't change that). If you
do get a 500 error (which is still likely in high traffic times)
please let me know here, so we can adjust this.

kthnxbai!

--
Matt Read
http://mattread.info
http://mattread.com

rick c

unread,
Nov 18, 2009, 6:38:52 PM11/18/09
to habari-dev
Is it time we take a look at another issue tracker, again? Maybe
indefero ( http://www.indefero.net/ )?

Rick

Realpolitik

unread,
Nov 18, 2009, 6:55:47 PM11/18/09
to habari-dev
Maybe we could redirect some of the trac traffic that's just for
downloading to a cached, static site?

On Nov 18, 4:52 pm, Matt Read <m...@mattread.com> wrote:
Reply all
Reply to author
Forward
0 new messages