Email Hostname Certificate Mismatch Causing sidekiq Queue Overload, Severe Site Instability

I tried sending a message from inside the Discourse Docker container using curl. Once I made sure to specify plaintext SMTP and port 25, I can send email via the host just fine:

$ cd /var/discourse/
$ sudo ./launcher enter app
x86_64 arch detected.
root@discourse-app:/var/www/discourse# curl smtp://172.17.0.1 --mail-from discourse@mydomain.com --mail-rcpt myname@gmail.com --upload-file README.md
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  7077    0     0  100  7077      0   575k --:--:-- --:--:-- --:--:--  575k
root@discourse-app:/var/www/discourse#

And this is what that test looked like in the host’s email logs:

May 24 16:53:49 localhost postfix/smtpd[25494]: connect from discourse-docker[172.17.0.2]
May 24 16:53:49 localhost postfix/smtpd[25494]: EB62CB5FCD: client=discourse-docker[172.17.0.2]
May 24 16:53:49 localhost postfix/cleanup[26008]: EB62CB5FCD: message-id=<>
May 24 16:53:49 localhost opendkim[1365]: EB62CB5FCD: can't determine message sender; accepting
May 24 16:53:49 localhost postfix/smtpd[25494]: disconnect from discourse-docker[172.17.0.2] ehlo=1 mail=1 rcpt=1 data=1 quit=1 commands=5

Given that I have specified no TLS and port 25 in my app.yml, and this worked until the rebuild yesterday, it’s looking more and more like the latest Discourse is ignoring my SMTP configuration in app.yml.

2 Likes