2009/11/11 Martijn de Munnik <mart...@youngguns.nl>: > remote host. Does anybody use a check which checks the complete mail loop? > I was thinking of sending a mail from a remote host (with webmin) to a test > mail account and see if I can download the mail with imap and then with pop > which removes the mail. The test mail account should also send a reply to > the original sender (maybe explaining it's a test address) and the test > server should also check for this reply.
This sounds a little elaborate to me, actually polling the mailbox via pop/imap, but it's comprehensive if nothing else. I should note that I only touch webmin very rarely; I find it curious that it'd have some sort of testing/probing functionality built in. Seeing as it's an administration tool, I suspect it's not the best tool for the job (but hey, if it's already there and it works, stay with it). We don't have an end-to-end monitoring setup, but I can think of something that should work. We use Nagios, which is free and flexible (but it can get a bit complex, and there's a learning curve). We have checks on a few points: * connect to the port and check the banner * check the length of the mailqueue on the machine This works well for us and catches most problems, but it wouldn't quite work for your scenario which is end-to-end latency-sensitive. I can think of something that would probably work with Nagios though: 1. inject an email periodically, directed to a special testing address (say, every 5min, via cron) 2. this should pass through the rest of the system like regular mail 3. at the "end" of the mail flow, pass the mail through a script (piping the mail out from /etc/aliases would do the job) 4. the script submits a "passive" check result to nagios (NSCA). this could be as simple as simply "freshing" the check, or you could analyse the headers for timestamps and look for a delay. 4a. you could turn this into an "active" check by having the script touch a file, then use an NRPE check to alert people if the file is too old.
