sendmail quits - cannot bind: Cannot assign requested address

scott_r_wagner · 08-21-2003, 11:55 AM

I am (or was) running sendmail 8.11.6 on a 2.4.20 kernel (Red Hat Linux 7.3). The installation has been running without problems for many months. A few days ago, upon reboot, sendmail began failing with the following entries in maillog:

Aug 21 12:12:40 mymachine sendmail[2421]: starting daemon (8.11.6): SMTP+queueing@01:00:00
Aug 21 12:12:40 mymachine sendmail[2421]: NOQUEUE: SYSERR(root): opendaemonsocket: daemon MTA: cannot bind: Cannot assign requested address
Aug 21 12:12:40 mymachine sendmail[2421]: daemon MTA: problem creating SMTP socket
Aug 21 12:12:45 mymachine sendmail[2421]: NOQUEUE: SYSERR(root): opendaemonsocket: daemon MTA: cannot bind: Cannot assign requested address
Aug 21 12:12:45 mymachine sendmail[2421]: daemon MTA: problem creating SMTP socket
[The above repeated at 5 second intervals 10 times]
Aug 21 12:13:30 mymachine sendmail[2421]: NOQUEUE: SYSERR(root): opendaemonsocket: daemon MTA: server SMTP socket wedged: exiting

Has anyone seen anything like this? This is sort of "out of the blue"; all other services are working OK, and the only recent change is a NIC replacement (previous NIC damaged during blackout.) I'm willing to send more info if any sendmail experts out there would like. Thanks!
-Scott Wagner, Rochester NY US

scott_r_wagner · 08-22-2003, 12:47 PM

I found the answer to this one - it seems my eth1: interface, which acqures its ip address via DHCP, conflicted with an entry:
O DaemonPortOptions=Port=smtp,Addr=a.b.c.d, Name=MTA
in my /etc/sendmail.cf .

When I manually fixed the DaemonPortOptions entry in sendmail.cf, sendmail worked fine.

I still have to figure out how to keep the DaemonPortOptions entry synchronized with the (DHCP-assigned) ip address.

smccrory · 09-14-2003, 10:02 AM

Scott, do you update a DNS name after you get a new one from DHCP? I.E. do you use DynDNS or similar service? If so, you should be able to use that named address instead of a numbered one. The only problem with that approach is that it may take a few minutes until sendmail will start correctly while the address gets the update and is propogated to your upstream DNS provider(s). In fact, I hourly retart sendmail in a cron.hourly script for that reason and it seems to help. Perhaps there's a better solution?

ehdwuld · 09-14-2003, 11:12 AM

while working in the now defunct telecom market,
It was decided that the best practice was to point customer circuits to the
interface rather than ip address

this may be a solution for your troubles

ehd

smccrory · 09-14-2003, 03:40 PM

Quote:

Originally posted by ehdwuld
while working in the now defunct telecom market,
It was decided that the best practice was to point customer circuits to the
interface rather than ip address

Can this be done with sendmail.cf?

ehdwuld · 09-14-2003, 05:40 PM

I don't know

ehd