No SMTP servers were available for mydomain.com. No hosts to try.

[Salvogiarracca]

Hello everyone, I have some problems setting up Postal as SMTP relay.

A bit of context:
I bought a VPS and I installed Postal on it, configured dns entries as the guide suggests, and finally having a 10/10 score on mail-tester.com. Then I setup a mail server on my private home server (IMAP/POP3) that let me store mailboxes privately. The accounts on this mail server are on the form [email protected]. Also I installed a wireguard VPN on my private server in order to have VPS (10.0.1.2) and MailServer(10.0.1.1) on the same LAN(10.0.1.0/24). All the services runs in docker containers with network_mode: host, except for Postal that is in its default configuration. Up to here no problem at all, each container can reach all the others (ping, telnet).

The problem:

• Created a new SMTP endpoint, with hostname myserver.com (the mailserver), port 25, ssl auto.
• Created a new route [email protected] associated to the previously created SMTP endpoint (myserver.com).

I suppose that this should be translated like this: "When I (postal) receive an email with the TO field equal to [email protected], i redirect it to the SMTP endpoint (myserver.com)". Sad story, it does not work.

The only useful information from UI is a strange SOFT FAIL:
No SMTP servers were available for mydomain.com. No hosts to try.

I don't have DNS problems because, as said before, each container can reach the others. I use a FQDN for the mailserver (mapped in /etc/hosts file). I don't think it should be a problem related to the mailserver because telnet command at port 25 works, but it is like the routing rule is not considered at all. I tried to add smtp_relays in the postal.yml, although I didn't understand how this is related to the SMTP endpoint in the UI. Even the option use_local_ns_for_domain_verification: true did not make any change.

I hope that I have been clear, any suggestion/help to fix the problem would be very appreciated.

[willpower232]

No hosts to try.

This is the biggest clue, I think this means that either there are no working MX records for the domain you're trying to send to or your server provider is blocking your SMTP connections outbound.

postal/app/senders/smtp_sender.rb

Lines 46 to 47 in fd3c7cc

	if @endpoints.empty?
	r.details += " No hosts to try."

[willpower232]

Ah potentially, I'm not really familiar with the low level network specifics but at this point, I imagine the problem will be something complicated like that.

[Salvogiarracca]

I will check that. Can you tell me how the SMTP endpoints (UI) and the smtp_relays (postal.yml) are related? I tried to add manually the smtp servers under the postal section in this way:
smtp_relays: [ 'smtp://myserver.com:25?ssl_mode=auto', 'smtp://10.0.1.1:25?ssl_mode=auto' ]
I assume it is syntactically correct because postal does not complain when restarted.

[willpower232]

If you have relays added to the config file, Postal will use them to send all outgoing messages so might actually be useful for your use case.

Endpoints in the UI are for passing on some email based on the rest of the route. You can also have non-SMTP endpoints to direct the message body at a HTTP API for example.

[Salvogiarracca]

I solved the issue. The problem was with DNS resolution. While /etc/hosts was correctly configured and telnet could reach myserver.com, it seems that Postal does not handle IP addresses well (the endpoint variable was empty). I decided to use a public FQDN instead, and the logs showed that Postal attempted to reach that host (unsuccessfully, as it was not an SMTP server). However, this time, the endpoint variable was printed correctly.

To address the issue, I installed dnsmasq and configured systemd-resolved to use dnsmasq as the upstream server. This setup ensures that only the DNS queries for the VPN domain are sent to dnsmasq.

Thank you for the support!

[willpower232]

Ah good stuff, the hosts file does behave in mysterious ways but then I guess its a little too easy to use!