[Tracker-discuss] Roundup installed
Erik Forsberg
forsberg at efod.se
Mon Nov 13 07:10:36 CET 2006
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Izak Burger <izak at upfrontsystems.co.za> writes:
> Hi all,
>
> A bit more research eventually showed that not only is perchild
> experimental and sometimes non-functional, it was also removed from
> apache2.2. Several other solutions have been written by others, for
> example:
>
> http://home.samfundet.no/~sesse/mpm-itk/
> http://mpm.metux.de/index.php/Main_Page
> http://www.telana.com/peruser.php
>
> Two of them keeps a pool of apache processes running under each uid, and
> the other runs as root for as long as it can before dropping privileges
> to the right user. Each has some problems of it's own.
>
> So I thought a bit more about the idea of keeping a seperate pool of
> processes running under the right uid. This led me to the idea of
> running a second apache under the roundup uid (that implements the
> separate "pool of processes") and let the frontend apache pass queries
> to it. Then I figured, if I do that I might as well use roundup-server.
>
> I've changed the configuration now to run roundup-server, with apache in
> front of it. Apache will serve all static content, and pass the rest to
> roundup-server. The idea is that we let apache do what it is good(and
> fast) at.
I don't quite see why this was necessary? What was so wrong with the
mod_python approach that it needed change?
The people on the roundup-users mailing list all seems to agree that
the mod_python frontend is the one with the best performance. It also
has the advantage of keeping the list of processes that need
monitoring at a minimum.
\EF
- --
Erik Forsberg http://efod.se
GPG/PGP Key: 1024D/0BAC89D9
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.2 (GNU/Linux)
Comment: Processed by Mailcrypt 3.5.8+ <http://mailcrypt.sourceforge.net/>
iD8DBQFFWAxcrJurFAusidkRAqOWAJ9ifPkpoDWsI5HgDH/bXJ+yW0kFWgCgjL/J
kMplN72I1Kuby8AW1xfDqDA=
=OJPI
-----END PGP SIGNATURE-----
More information about the Tracker-discuss
mailing list