On 2017-08-17 Micah Garlich-Miller wrote:
I forgot to mention: the timestamps on your problem emails happen before they started work on the ironports. So I don't think it's related.
DATA
<<< 550 trevor@tecnopolis.ca... 451 4.7.1 Greylisting in action, please come back later
Thanks guys, Gilbert was right, the problem was on my side. I upgraded to Fedora 26 2 weeks ago, and that's the only time sendmail and milter-greylist changed, but I thought, nah, this couldn't have been broken for a whole 2 weeks and I'm just hearing the complaints now! Dropping email for 2 weeks... don't even want to think about that...
But, a quick telnet port 25 and some commands later, yup, it's me. Looks like a bug that hit in 2015 and was supposedly fixed but is now regressed. It could be a Fedora thing, as it looks like it's a missed patch because Fedora is still using sendmail 8.15 instead of 8.16 (why?) which has the official fix.
https://bugzilla.redhat.com/show_bug.cgi?id=1482808
The quick fix: dnf downgrade sendmail
Now gives me correct: 451 4.7.1 Greylisting in action, please come back later instead of buggy: 550 validuser@problemsendmailhost.com... 451 4.7.1 Greylisting in action, please
And yes, it looks like I'm once again the first one in the world using Fedora 26 + sendmail + greylist. Bah. Always the guinea pig. Oh well, it is my pleasure to serve, so all you CentOS/RHEL people can have your bug-free major releases ;-)
From my bz, as to why this wasn't immediately noticeable:
It's really insidious because email doesn't just stop: there are still 2 cases where emails get through: the tuple is in gm's hard or auto-whitelist, or the end user resends the email after the gm timeout period. So many emails will get through and mail admins may not know there is a problem until they start getting some phone calls about the sporadic missing emails...
Thanks to Adam for testing too.