Lighttpd

I disagree; if you want many forks to be running, you want to "watch" all of them (waiting till the last one dies before restarting doesn't make sense).

So you just have to use multiple supervice/spawn-fcgi instances. lighttpd has no problem balancing over multiple sockets.

Good fastcgi applications should have their own worker-fork/thread support if they can only on request at a time, so they can share some memory for common settings/caches/...

This is extremely impractical when working with daemontools, specifically, which is what the -n feature was supposedly added to deal with. It means that you have a bunch of almost-identical-but-not-quite configurations that supervise is dealing with and these all have to match up with the lighttpd configuration.

I could write something to fork a bunch of processes when -n is in effect, but they wouldn't all be on the same socket. Honestly, I think that -F should do what it is advertised to do and when -n is in use, it doesn't.

Obviously you didn't get the message...

IT DOESN'T MAKE ANY SENSE

1.) You don't need to be rude.
2.) It does make sense. I am describing a real-world problem.
3.) We don't always have the luxury of controlling the source for the processes we're trying to launch.
4.) The fact is, spawn-fcgi lies about what -F does; namely, it doesn't do what it advertises when -n is used.

According to the provided help, -n is supposed to keep the process from going into the background specifically so that this works with daemontools. -F is supposed to fork multiple processes on the same socket.

-F does not work in the presence of -n because someone made a poor choice.

I don't know how to say it any more simply than that.

Can you please read again my first post...?

If you want many forks to be running, you want to "watch" all of them (waiting till the last one dies before restarting doesn't make sense)

I don't know how to say that any more simple... perhaps a little bit more complex?

spawn-fcgi doesn't supervise your forks, and doesn't restart them. So if you fork 10 workers and 9 die, you have only one remaining and your supervise wouldn't care. So you completely screwed the idea behind supervise. That is just stupid, and it is therefore a good choice not to support it, even if the help of spawn-fcgi does not indicate that. I myself would just kill the -F option.

If you really don't care about that and want to screw your setup, just use a wrapper script:

#!/bin/bash

TARGET=/your/fcgi/app
N=10
declare -a pids
for((i=0;i<N;i++))
do
    $TARGET &
    pids[$i]=$!
done

for((i=0;i<N;i++))
do
    wait ${pids[$i]}
done

If you want to discuss it further, join us in irc.

But do not again ask us to provide ugly workarounds for your problems. We have enough problems with existing bad code.

missing feedback.

See http://cgit.stbuehler.de/gitosis/multiwatch/about/