[Catalyst] Notes on installing lighttpd and FastCGI for Catalyst

Bill Moseley moseley at hank.org
Wed Mar 8 16:12:48 CET 2006


On Tue, Mar 07, 2006 at 11:01:22PM -0800, John Wang wrote:
> On 3/7/06, Bill Moseley <moseley at hank.org> wrote:
> >
> > What the advantage of running multiple FastCGI process managers on the
> > same machine?
> 
> 
> I'm not sure there is any. A few RoR posts had lighty calling multiple
> FastCGI sockets but perhaps those are tied to a single FastCGI process a
> piece, I didn't look too closely. I'm not sure if it's a worthwhile
> configuration but included it just for completeness. If there's no reason
> for it, I'll remove it.

Don't remove it -- the more examples the better.

I wish I understood more of this.  For example, what does defining
multiple servers do in lighttpd?  When a request comes in how does it
decide which socket to use?

Is it possible that there's overhead in the process manager when a new
request comes in so that running multiple servers (even on the same
machine) mitigates that at bit?

Under forking Apache more processes are started as needed.  But,
IIRC, the process manager has a static number of processes running.
Does this mean you should start more servers than you normally need
for the inevitable slashdotting?

Again, on Apache (and maybe lighttpd?) when using the server's
process manager ("static" mode) it's the same case -- a fixed number
of application processes (if I remember correctly).


> My primary interest is in the load-balanced TCP config which isn't covered
> in the C::E::FastCGI POD at the moment, but perhaps it can be included.

Good idea.  Under Apache I assume it's just a matter of defining
multiple FastCgiExternalServer configs.


-- 
Bill Moseley
moseley at hank.org




More information about the Catalyst mailing list