A checklist that keeps a mail server alive

Mail servers do not fail dramatically. They fail quietly: a queue grows, a certificate expires, a backup has been “successful” for eight months while writing empty archives. The checklist below is the one I actually use, and every item on it exists because I once skipped it.

Daily

Look at the queue depth, not just the service status. A running Postfix with four thousand messages in mailq is a broken server that answers to systemctl. Check the last delivery timestamp per domain you care about — silence is the symptom.

Weekly

Verify that outbound mail leaves and lands. Authentication results, not “it was accepted”. A message accepted by your relay and rejected by the destination is still a failure, and the bounce may be going to a mailbox nobody reads.

Monthly

Restore one file from backup and open it. Not “the job exited zero” — actually open the archive. Also check the certificate expiry with enough runway to renew calmly rather than at 23:40 on a Sunday.

Once, when you set it up

Write down where the DKIM keys live, who owns the domain, and what happens if the machine is rebuilt tomorrow. The rebuild is the test. If the answer requires archaeology, the documentation is the outage.

None of it is interesting work. That is the point: interesting work is what you get to avoid.