[enterprise] RCPT TO Stage Best Practice for Split Delivery with Catch-All/Sub-Addressing Support

[dennisoderwald]

We have configured Stalwart and it is now operational. However, we are unsure whether this is a bug or intentional behavior:

We have Google Workspace as our incoming MX and forward emails from unknown accounts (i.e., in Google) to Stalwart. Outgoing emails that do not exist locally (in Stalwart) as accounts are sent via Google's SMTP relay. This works so far.

The next step was to ensure that features such as catch-all and sub-addressing were working. Although the values were set to “true,” they were not working.

Only after we created three rules in the RCPT TO stage as recipient rewrites did this start working properly.

We are now asking ourselves whether the technical implementation complies with best practice and will remain compatible with further updates, or whether there is a better recommendation.

Outbound Settings:

Rule 1:

is_local_address('', rcpt) ) => 'local'

else:

'relay'

Inbound Settings (RCPT Stage):

The Recipient Rewriting as text:

Rule 1:

is_local_address('', rcpt) => rcpt

Rule 2:

contains(rcpt, '+') => split(rcpt, '+')[0] + '@' + split(rcpt, '@')[1]

Rule 3:

!is_local_address('', rcpt) => rcpt_domain

else: false

[mdecimus]

Your settings are correct and will remain compatible with future versions.
The issue you experienced with sub-addressing is because this setting only works with internal recipients, please track enhancement #2217 which, among other things, will add support for sub-addressing in split-delivery setups.

[mdecimus]

We currently have v0.13.4, and I have noticed that emails to the catch-all are not arriving. Has anything changed in this regard with the latest version?

Which version were you using before? If it was >= 0.13.0 then nothing has changed in that area.

In the “Delivery Attempt History,” the email loops at least when I try to send something to a catch-all...

Try setting the log level to trace and check your logs for more details.

[dennisoderwald]

@mdecimus Sorry for the late reply. Attached is a log that will hopefully help resolve the situation. If you need it unedited, i.e. without the domains being obscured, I would send it to you by email.

stalwart.log

[dennisoderwald]

Which version were you using before? If it was >= 0.13.0 then nothing has changed in that area.

I can't answer that off the top of my head, which version we had in use before...

[mdecimus]

Looking at the logs it seems that you are using Stalwart as a relay host? For example this line:

from = "[email protected]", to = ["[email protected]"]

This might be the reason of the routing loop. Make sure that all local domains are defined in Stalwart, and that if you are indeed relaying, that the remote host does not return the relayed message back to Stalwart.

[dennisoderwald]

@mdecimus Thanks for your reply. Let me explain our setup to you, and maybe it will help you understand where my mistake in thinking or application lies.

We primarily use Google Workspace (for outgoing and incoming emails), but there are email accounts (via LDAP Integration) that only exist in Stalwart, and these should be sent via Google Workspace. Google Workspace is configured to forward unknown accounts to Stalwart, and this works fine. The only problem is that if an entry does not exist in either Google or Stalwart (catch-all case), the email runs in a loop even though a catch-all should actually be accepted by Stalwart. So, in principle, anything that is not found in Google should remain in Stalwart in the catch-all case. Of course, “@domain.de” is defined as an alias for this purpose.

Ultimately, we want to use sub-addressing as well as catch-all in Stalwart. Outgoing emails should run via Google if the accounts are not known to Stalwart. With catch-all, it basically has to loop via Google Relay, because otherwise it doesn't know which accounts are to be delivered locally and which externally. This is because Stalwart only knows its own accounts, not those from Google.

I hope the case is understandable so far. I have mapped out the current configuration for you here once again.

[mdecimus]

I actually thought that everything was working now, but with some recipient addresses, despite catch-all, I get an error when sending from locally:
recipient address [email protected] not accepted by the server
server message: 452 4.4.5 Rate limit exceeded, try again later.

Looks like a configuration issue but since you deleted this comment I assume it's now fixed?