Grepular

My primary MX is a server in my house. My secondary MX is a Digital Ocean VPS:

grepular.com mail is handled by 10 mx1.grepular.com.
grepular.com mail is handled by 20 mail.grepular.com.

I went on holiday recently and whilst I was away my primary MX had a kernel panic, so it was “offline” for about a week. Not a lot I could do about it because I need to be physically present to enter my full disk encryption key to reboot it anyway. My expectation was that my backup MX would start receiving all mail, and queue it up ready for me to get home and sort things out.

Nope… The primary MX was in an odd state where all of the services running on that box would accept TCP connections, but then there was no data transfer. This seemed to adversely affect several mail senders, including GMail, Sendgrid and Github. There are probably others, but those are the ones I noticed. None of them attempted to send mail to my secondary MX. It’s as though they connected to my primary, sat there waiting for a welcome SMTP banner until they timed out, and then either gave up, or waited to try the primary again later, ignoring the existence of my secondary.

There is definitely nothing wrong with my backup MX as I have other domains that use that host as a primary, and they had no problem receiving mail from those services. And after restarting the box, I put in a temporary IPTables rule to reject incoming port 25 connections immediately and then sent myself a test email from GMail. GMail, then successfully delivered to the backup MX at that point.

Frustrating. So if you’ve emailed me in the past week and not had a response, or got a bounce, now you know why.