Re: Debugging COMM:START-UP-SERVER hang

Martin spake:

> I suggest adding the -f option to strace (which will make it show all threads)
> and then try to make a new connection to the port to see if it is accepted.

Thanks Martin! Like I mentioned earlier, new connections weren't htting
the ACCEPT phase. There's a little tool `cgi-fastcgi -bind` which can mock
a FCGI process. That was failing at the CONNECT phase.

> Also, look at the output of netstat -antp (or ss -antp if you don't have
> netstat) to check that the process still has a listening socket and that the
> system doesn't have too many sockets stuck in strange states like CLOSE_WAIT.

ooohh... that sounds like it could be the reason. I'm not certain what to do
about that ... time to break out my TCP book again I suspect.

> You could also look at the output of lsof -p $pid to check that the process is
> not running out of file descriptors.

oh yeah... that also sounds plausible. It *does* take a good week for the
acceptor func to hang.

I've restarted the service again, so I'll try your suggestions in about a week
when it stops working again 🙁

Thanks Martin!

A.