Network timeout for IMAP, SMTP, NNTP
 

Hi,

I have a new installation of Debian Etch. I experience network timeout problems using a mail client to connect to our mailserver (tried 3 different clients, icedove (=thunderbird), evolution, balsa)) using IMAP or SMTP, but also when I try to connect to a news server on the internet. I can access the same mail server from my old machine, right next to me. The settings are the same as far as I can find out.

What works:

- internet connection (web surfing)
- "manual" smtp connections to the mail server (telnet port 25)
- "manual" imap connections to the mail server (telnet port 143)
- connecting a network share on the mail server (this is a Windows server)

I tried to search the log files for unusual entries, but didn't find anything which looks strange to me. I tried tcpdump to watch the communication between my machine and the mail server, there is one packet from me to the server and one packet back, then timeout. I have no firewall settings on my machine.

What did I oversee? There must be something very simple an obvious going wrong here, but I just can not find out. Please someone give me a hint!

Thanks,

Christoph Jäger

Sounds a bit like a dns problem. Did you telnet by IP? Can you ping to the relevant server by hostname?

Ping works, I also tried specifying the IP address directly in the mail clients, got the same result.

have you required authentication on the server or anything similar?

Of course to do IMAP I need to specify a username and password. But SMTP for instance works without any authentication (I can send emails when I simply telnet to port 25 and speak SMTP with helo, mail from, rcpt to, data, quit).

And nothing in the logs of either the client machine or server? Surely there must be something?

Hi Billy,

I did a "tail -f messages syslog daemon.log auth.log dmesg exim4/mainlog kern.log mail.* user.log" in the /var/log directory (just to be sure to not miss any log file), and there wasn't a single new entry from when I hit "Get Mail" in the email client to the timeout message. Also, I can not find log entries in the Event Viewer on the Windows machine. And nothing in the logs of the two firewalls between me and the mailserver (the server is located at a different office with a VPN connection in between).

Maybe I need to enable some more debugging to get some log entries / error messages? But where can I do this?

Sounds to me like a VPN issue

What MTA are you using

Hi Billy,

Thanks for your help, it a little bit lead me in the right direction. I found enough evidence for the source of the problem to finally solve it.

To make it short, I had to disable tcp window scaling (I suspect my Zyxel firewall to be the reason for this, but have not yet confirmed it).

Here is how I found out:

- used the default installed exim4, but as I don't have any experience with it, switched to postfix (as I have a little bit of experience with this).
- switched on debugging with postfix -> same connection problems as with mailclient talking directly to the server
- used tcpdump to log network packets.
- used wireshark to inspect packages, tried sending emails with my mailclient (which does not work) and manually with telnet on port 25 (which does work) to see what is different. Now suddenly I see a "TCP previous segment lost" error message just after I typed "helo HOSTNAME". With the manual method, I didn't wait for the server's greeting, just used the helo command. My mailclient waited to see the server's greeting, but this never arrived.
- typed "TCP previous segment lost" in Google, found this: http://www.linuxquestions.org/questi...d.php?t=458634
- sysctl -w net.ipv4.tcp_window_scaling=0 really solved my problem

Thanks again for your help,

Best Regards,

Christoph Jäger