F9 - sendmail won't send on startup?

Tim ignored_mailbox at yahoo.com.au
Mon Jun 23 08:45:03 UTC 2008


On Sun, 2008-06-22 at 11:17 -0700, Knute Johnson wrote:
> I've got two installations of F9 running, one x86_64 and the other
> i386.  They both fail to send email initially but if I restart
> sendmail they send fine after that. 

Just so you're not alone, I also have problems with sendmail, but not
like you've outlined.  There's a long story below, but the shorter
version is to see if turning off NetworkManager and using the network
service, instead, helps.

My sendmail apparently without protest, but it just doesn't send.
Logwatch mails rarely get through.  They didn't end up in /var/mail/root
(i.e. local delivery doesn't work), and when I put an alias for root
that sent mails to my working mailserver (on another computer, that
everything else on the LAN works absolutely fine with), they don't
arrive either.  Periodically, there'll be a postmaster bounce that does
arrive, complaining that it wasn't able to send the post (oddly those
postmaster mails seem to get there).

Looking through various logs, I see things are still insisting that
they're "localhost.localdomain" when they shouldn't be (I have manually
set my Fedora 9 network configurations for a domain, long ago, since the
install and network manager didn't pay attention to my DHCP server), and
I see plenty of "could not send message" notices, and some nameserver
timeout messages in the very tardy logwatch posts.  If I do a service
sendmail restart, the messages stuck in the queue are sent.

My LAN SMTP server is not firewalled against other LAN PCs, I have fully
functioning DNS (it responds, all machines resolve in forward and
reverse directions, it has MX records, etc.).  This Fedora 9 box is the
only one that has problems sending mail.

At the moment, my guess is that the network comes up too late, sendmail
gets upset, and doesn't recover.  While it apparently does try to send
again (I see warnings about being unable to deliver for several hours,
which suggest it's tried during that period), it doesn't seem restart
whatever tries to find the address.

NTP seems to be the same.  Tries to start, fails, never recovers.

The Fedora 9 box's /etc/resolv.conf file is almost correct.  It lists
both a domain and a search entry (you're not supposed to do that,
according to manuals I read ages ago), but the nameserver address is
correct.

Doing a "locate /etc/resolv.conf", I find several backups.  Some of them
have errors.  Perhaps they're being looked at during the problem?  You
might want to check for the same.  I've now edited all of mine to be the
same (having just one search line, and one nameserver line).  Rebooted,
and things weren't much better.

Next I deleted all those backups, configured the original files
correctly and rebooted.  Saw messages like the following (retyped):

After bringing up eth0
messages about cp cannot rename /etc/resolv.conf.predhclient.eth0 permission denied
mv cannot move /etc/ntp.conf to /etc/ntp.conf.predhclient.eth0 permission denied.
/sbin/dhclient-script line 399 /etc/ntp.conf permission denied
/sbin/dhclient-script line 402 /etc/ntp.conf permission denied

Next I turned off NetworkManager, started the network service,
configured the system to boot that waym, deleted all the backups,
manually set the originals to be correct, and things worked fine,
despite a daft error message while booting (much like the above).  

It'd be a lot easier to report these things if someone would get
boot.log working properly again (mine just shows some virtually useless
entries about DHCP).  Yonks ago, several releases back, someone killed
it, stating they'd come up with something better, but never did.

-- 
[tim at localhost ~]$ uname -r
2.6.25.6-55.fc9.i686

Don't send private replies to my address, the mailbox is ignored.  I
read messages from the public lists.







More information about the fedora-list mailing list