httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ricardo Cantu <rica...@smartcsc.com>
Subject Re: mod_fcgid
Date Tue, 22 Sep 2009 13:34:34 GMT
On Tuesday 22 September 2009 7:03:36 am Jeff Trawick wrote:
> On Mon, Sep 21, 2009 at 7:16 PM, Ricardo Cantu <ricardo@smartcsc.com> wrote:
> > I'm a long time user of mod_fcgi
> 
> BTW, I think you mean "mod_fastcgi."  I haven't found a true "mod_fcgi,"
> though some people on the web have used that to refer to mod_fcgid.

Yes, mod_fastcgi,

> 
> > and would like to start using mod_fcgid. I've
> > been running mod_fcgi with a custom patch of mine. I like to see if it
> > could
> > be included in mod_fcgid or maybe you can give me an other way to
> > accomplish
> > what I need without the patch. What I have is one C program that lives on
> > a linux server. I want to use a web browser as the UI. The C program is a
> > classic "one instance to one UI". If 5 people want to run this program I
> > need
> > five separate and persistent copies of the program running, one for each
> > person. When one is done then the program exits.
> > What I've done is sym-linked my program to make unique program names
> > like:
> > program-1
> > program-2
> > program-3
> > etc..
> > Then each browser asks for a different name.
> > plus the patch that allows setting  killInterval to 0, so the process
> > manager
> > won't kill my process ever.  Here are the settings:
> > -maxClassProcesses 1 -singleThreshold 1 -killInterval 0.
> >
> > I know I could have done it with FastCGIExternalServer, but I wanted
> > everything to be dynamic.
> 
> I think this may work for you:
> 
> DefaultMinClassProcessCount 1
> DefaultMaxClassProcessCount 1
> IdleTimeout 2147483647
> ProcessLifetime 2147483647
> 
> Those magic values for IdleTimeout and ProcessLifetime are as close to
> "unlimited" as you can get with the current code.
> 
> Note that these can only be set globally with mod_fcgid.  That's probably a
> much bigger problem than the "unlimited" hack.
> 

It's seem very reasonable to have "0" (unlimited) as valid value for 
killInterval, IdleTimeout, and ProcessLifetime. As I don't think the actual 
time of zero seconds would make sense in any of those variables. 

Any objections to such a patch?

Mime
View raw message