SMTP server is slow to connect

[kechols]

Lately, I've been receiving a lot of complaints about it taking a while for email to leave the outbox from my endusers. It's taking about 18 seconds to connect to the mailserver (according to our network monitoring server). It's been getting more noticable over the last week or two, so far I have seen no more than 65 SMTP threads (from doing ps auxw |grep -ic smtp) and from the server (telnet localhost 25) it is immediate. This is the only service that has this problem, all the other services respond in less than 1second or I would think it might be a network problem.
This is happening on zimbra 3.0.1 GA. I will be happy to provide any logs or configs that might be helpful.

Thanks,
Kyle

[soxfan]

I guess the first question would be why are you running such an old version of Zimbra? If possible, I would suggest that you look into upgrading to the latest version (4.5.6).

[kechols]

The last upgrade (to version 4.5.0 i think) failed, badly, I restored from backup and then had to spend the next week fixing email accounts as they called in. Because for some reason, some of the emails corrupted and wound up delivering the same emails over and over again. Since that happened, my boss has been learing of Ok'ing another upgrade attempt. Regardless, This is a problem that seems to be getting noticed by more people everyday and any help would be greatly appreciated.


Thanks,
Kyle

[soxfan]

I can certainly understand the reluctance to try another upgrade based on your previous experience. However, there have been so many bug fixes, performance improvements, new features, etc. released since version 3.0.1, and there is a good chance that the problem you are experiencing has been addressed.

I guess if upgrading is not an option try looking in /var/log at the messages and maillog files. Also, in the current version of Zimbra /opt/zimbra/log/mailbox.log contains a lot of information. However, it may have been called something different, and may even reside in a different location with the version you are running.

[bhickey]

I have to agree. . .3.0.1 is ancient. You really need to upgrade.

I understand how a bad upgrade experience can make one gun shy, but to even receive good advice from the forums, you should consider moving to a newer version. My guess is that very few are people still running version 3.

[kechols]

Oh I'd love to, I've been pushing to upgrade since I took over the server, over a year ago. The first attempt took 6months + to get an ok, and a recommendation from a Zimbra salesperson. I did manage to get an Ok to go to 3.1 this weekend or next (whichever I prefer). But it's something where I'm going to have to let it run for 3 or 4 weeks, then try to upgrade again, until it's current.

The thing that really bugs me about it, is after the ~18second handshake I get the zimbra banner (when telneting) and the email goes out (via outlook), no delays, other than waiting on the zimbra banner. So I'm really just trying to figure out what could be causing the handshake to take so long.

I know it's an old version and appreciate any ideas that might help in figuring it out.

Thanks again,
Kyle

[esoleyman]

Oh I'd love to, I've been pushing to upgrade since I took over the server, over a year ago. The first attempt took 6months + to get an ok, and a recommendation from a Zimbra salesperson. I did manage to get an Ok to go to 3.1 this weekend or next (whichever I prefer). But it's something where I'm going to have to let it run for 3 or 4 weeks, then try to upgrade again, until it's current.

The thing that really bugs me about it, is after the ~18second handshake I get the zimbra banner (when telneting) and the email goes out (via outlook), no delays, other than waiting on the zimbra banner. So I'm really just trying to figure out what could be causing the handshake to take so long.

I know it's an old version and appreciate any ideas that might help in figuring it out.

Thanks again,
Kyle

Kyle,

I recommend that you obtain a good packet trace at both the desktop (email client) and server to better understand why the 18 second delay is occurring.

On the server, you can run the tcpdump command by giving "/usr/sbin/tcpdump -s 1500 -i EXTERNAL_INTERFACE port smtp -w /tmp/server_trace.pcap" and view it on the desktop using Wireshark (Wireshark: The World's Most Popular Network Protocol Analyzer)

On the desktop, you can run Wireshark to do the graphical equivalent of tcpdump.

Using both traces, you can now find out where the 18 seconds delay is occurring given the view from both the desktop and server.

I hope this helps.

-
Emil Soleyman

[kechols]

I found the problem, it was related to a reverse dns entry made months ago, just not sure why it just started causing a problem. Thanks for the help