mailq - indicates problems?

[Fred Martin_1]

My mailq remains mostly empty, but lately and especially after upgrading to sendmail 11.1, I get many queued messages with the following messages - to many different destinations, not just the ones shown:

(host map: lookup (adelphia.net): deferred)

(Deferred: Connection timed out with adelphia.net.)

Occasionally a "denied" shows up too, I don't have an example for you. Odd thing though, the "denied" message gets delivered OK with a later processing of the queue. So it's not the message that's the cause.

The issue seems temporary, in all cases the mail is later delivered.

Any ideas about what the cause could be?

[Jeff Schussele · Accepted solution]

Hi Fred,

There can be several reasons for this including:

Destination server load
DNS not available
DNS lookup timeouts
Connections dropped mid-stream

I suspect most of these are due to problems on the destination end.

Rgds,
Jeff

[Steven E. Protter]

This problem is common with my servers that route mail directly onto the public Internet.

It is commonly caused by delahys in resolving DNS requests.

It can be caused by the following problems:
1)Incorrect DNS configuration on the receiving server
2)Internet congestion prevents timely resolution.
3)Local DNS server problems causing delays in resolution.
4) Unreliable local Internet Access

Of the above problems, you can only deal with #3. Run sone nslookup tests, for mx as recommended in your other thread which I posted to a few times. The DNS server there may be suspect and may not be handling mx requests correctly. The same methodology can help you work on solution #3.

As far as #4 goes How reliable is yours ISP?

SEP

[Fred Martin_1]

Well, a lot of different destinations are involved; not sure that's it.

However, I have recently had suspicions that my internet connection (144K DSL) is being stressed though. That could be it; we started with only 10-15 internet users and now have many more.

[Fred Martin_1]

I'm convinced now that the issue is I've got too much traffic for a 144K pipe. As an example, I put a ping on one of the hosts on my ISP's network, just on the other side of the pipe. At the same moment the ping begins to show poor return times from the host, several outbound messages show up in the mailq. A look at the files in the mailq with 'vi' shows DNS time out errors.

This also explains why I had so much trouble when I sent sendmail up to deny email from domains that could not resolve; DNS was working overtime and timing out, so I was denying valid email.

[Geoff Wild]

Don't know if this will help....

Try increasing the q time on sendmail:

If yours is:

sendmail -bd -q15m

Try sendmail -bd -q1h

Also, for DNS, add this to your /etc/resolv.conf

retrans 2500
retry 2


From man:

The values for retransmission timeout and number of retries to be attempted can be configured. These correspond to the retrans and retry fields in the _res structure. The following three options, listed in the order of precedence, have been provided for configuring the retransmission timeout and retry values.
The retrans is to be specified in milliseconds, and its default value is 5000 milliseconds. The default value for retry is 4.


Rgds...Geoff

[Scott Donaldson]

Hi,
The "(host map: lookup...)" message is similar to what I experienced during a brief experiment with sendmail.org's 8.12.8 on a 'selectively patched' 11.00 box. I noted that I was also collecting a number of messages in the queue with that error. I was able to telnet to the primary mx for the destinations and found almost universally that it was a Cisco PIX firewall at the other end....

After doing more research I came to the conclusion that my IP stack must be busted (I'm guessing PMTU discovery) and went searching for patches. Sure enough I found PHNE_26771 and applied it. It rebuilds the kernel and upon reboot all the messages were sent! I double checked the logs to make sure!

FWIW

Scott.

[Fred Martin_1]

Scott,

That's worth looking into, thanks very much.

Fred

[Fred Martin_1]

PHNE_26771 shows up when I do an swlist so I believe I have that patch in already...