Running Panther Server: First nightmare with Cyrus

Posted by Pierre Igot in: Macintosh
February 16th, 2004 • 12:42 pm

Well, I guess it had to happen sooner or later… After a fairly smooth experience setting up our new Xserve with Panther Server as a mail and web server 10 days ago, I just experienced my first major troubleshooting nightmare.

All of a sudden, I became unable to check my email on the server (POP). The server would simply not respond. All other services were still running fine and, based on what Server Admin was telling me, the Mail services were running fine as well.

I was still able to send email via SMTP on the same server, and the web server was running fine.

I first suspected a misbehaving firewall, and double-checked all my rules. But POP server traffic was clearly allowed to flow freely.

Of course, all this happened just an hour before I was supposed to leave home for a 3-hour trip. I didn’t have time to do much more troubleshooting, and just took the PowerBook with me with a view to exploring the issue further later on in the afternoon. (Since all this was still fairly new stuff, I hadn’t got around to installing the Server Admin tools on the PowerBook yet, so I had to do that first. Fortunately, this aspect of things is as simple as installing any other piece of software.)

I ended up spending the whole evening on the problem, but I finally fixed it.

After spending more time fiddling with the firewall settings and checking the logs, I noticed all kinds of unexplained traffic coming and going to the 127.0.0.1 address, which I do not use for this server. I also noticed traffic going to 224.0.0.x addresses. All this looked rather suspicious to me, so I started fearing that my mail server had actually already been blacklisted because of insufficient security. I looked for a site that would be able to tell me if my domain was blacklisted, and found this one. But my domain didn’t appear anywhere. Besides, the SMTP part of it was still working, so that would have contradicted the blacklist theory.

I still created a few additional rules blocking the traffic to and from these suspicious-looking addresses. I hope I didn’t disable block important in the process. The user interface for the firewall in Panther Server is not exactly super-intuitive. And I don’t exactly have tons of time to learn about all this.

I was basically back to square one when it came to my problem, though. The POP server was still not responding. That’s when I finally thought of checking the log for POP services, and noticed stuff about a “fatal error” in the mail database, asking for “recovery”. Sometimes it’s hard to tell what in a log file is important and what is not, but this definitely did not look good.

I then tried to search the Panther Server documentation for database recovery. I found something in the document about Mail Services, but it wasn’t clear enough. I searched Apple’s support web site for Mac OS X Server, but only found something relevant to 10.1 and 10.2. Nothing on recovering the mail database in 10.3. I guess they haven’t had time to update their knowledge base yet!

I ended up searching the Apple Discussions site for Panther Server and finally found a post that was precisely about the same problem. Phew!

I followed the instructions carefully (which also required that I use SSH to connect to the Xserve in the Terminal, something that I had never done before, but was fortunately not as tricky as I had feared). I restarted mail services. And miracle of miracles… The POP server was working again!

Now the big question for me is: Why did the mail database become corrupted so easily? I have only set up a couple of email accounts on the server, and have only used them for testing purposes for about a week. Not exactly intensive usage! We are about to move all our staff (approx 15) to this server, creating an account for each member. I sure hope that this kind of problem is not going to be a regular occurrence! At least now I know what happened and how to fix it… But still! It was a pretty scary experience.


One Response to “Running Panther Server: First nightmare with Cyrus”

  1. Pierre Igot says:

    The thread about the problem on Apple Discussions is no longer there (don’t know why Apple can’t leave these threads up at the same address permanently), but here’s a link to the new tech note that clarifies the problem and provides the fix:

    http://docs.info.apple.com/article.html?artnum=107996

Leave a Reply

Comments are closed.