Re: One more bug

[Jan-Henrik Haukeland]

Rory Toma <address@hidden> writes:

> Actually, the sympton was the reverse. Monit would die just fine (all
> threads) but the socket hadn't timed out yet. So, when monit_http ran,
> it would find the socket in use, and exit. However, the rest of the
> threads would keep running.

Aha, this is another problem. In this case it's the OS's "fault". If
there is data on a socket/port the OS may choose not to release the
port for some time (and prevent other sockets to bind to the port;
google for TIME_WAIT). We use the SO_REUSEADDR socket option on the
server socket (man 7 socket) so the OS should allow the program to
bind to the port even if there is activities on a previous socket
bound to the same port and that's really (almost) all we can to from a
programs viewpoint.

-- 
Jan-Henrik Haukeland