Re: Postfix with Cyrus Help

["nate" <aphro@aphroland.org>]

Curtis Vaughan said:

> Here is the message I get.
>
> Pacifica:/etc/postfix# tail -f /var/log/daemon.log
> Oct 19 12:41:38 Pacifica inetd[13228]: time/tcp: bind: Address already
> in use
> Oct 19 12:41:38 Pacifica inetd[13228]: daytime/tcp: bind: Address
> already in use
> Oct 19 12:41:38 Pacifica inetd[13228]: discard/tcp: bind: Address
> already in use
> Oct 19 12:51:38 Pacifica inetd[13228]: auth/tcp: bind: Address already
> in use
> Oct 19 12:51:38 Pacifica inetd[13228]: imaps/tcp: bind: Address already
> in use
> Oct 19 12:51:38 Pacifica inetd[13228]: pop3/tcp: bind: Address already
> in use
> Oct 19 12:51:38 Pacifica inetd[13228]: imap2/tcp: bind: Address already
> in use
> Oct 19 12:51:38 Pacifica inetd[13228]: time/tcp: bind: Address already
> in use
> Oct 19 12:51:38 Pacifica inetd[13228]: daytime/tcp: bind: Address
> already in use
> Oct 19 12:51:38 Pacifica inetd[13228]: discard/tcp: bind: Address
> already in use


ok, kill inetd again, and verify it is down by checking the
process list (ps auxw | grep inet)

next you have to do a bit of waiting and some netstat commands
(or reboot if your in a hurry and don't care about downtime):

the services are declared in /etc/services, so to find the port
that daytime uses:

grep daytime /etc/services

which tells me it uses port 13

then run

netstat -an | grep 13

if you see results come back which have port 13 in them, and on
the far right(last column) have the word TIME_WAIT then you have
to wait until that goes away(keep doing netstat) before trying
to start inetd again.

if you see on the far column the word LISTEN that means something
else is using that port. if the far left says tcp then do:

fuser -n tcp 13

(replace tcp with udp if far left says it is udp) to find the
process id of the program that is using that port. then find
it in the process list (ps auxw | grep <PID>) where <PID> is
the number that fuser returns.

I would expect it to be another instance of inetd..

repeat with all the other services that came up with "address already
in use" until running netstat for each one returns nothing, no
LISTEN, and no TIME_WAIT or anything related to those ports.

then start inetd again, it should be back to normal.  Note that
waiting for TIME_WAIT can take some time, I have seen it take up
to 20 minutes for the system to release the socket/port.

good luck!

nate