Context

Using the recommended volume configuration won't save your mail queue from restarts:
https://github.com/tomav/docker-mailserver/blob/f225e14a219824e4ebb34e63b144e6d7c1314bcb/docker-compose.yml.dist#L13-L17

What is affected by this bug?

Deferred mails in queue will not be resent after a reboot or container rebuild.

When does this occur?

When some mail gets deferred and queued and, later, you reconfigure your container (quite common for fixing some kind of problem, like a relay password that has changed), the mail queue is lost.

How do we replicate the issue?

1. Configure with SMTP_ONLY=1 and some DEFAULT_RELAY_HOST that makes the sent mails fail. The failure might be a wrong password or (in the case below) a wrong TLS setting.

2. Send a mail.

3. Execute postqueue -p in the container. You'll see an output similar to:

   $ docker-compose exec smtp postqueue -p                                                                                                                                                         
   -Queue ID-  --Size-- ----Arrival Time---- -Sender/Recipient-------                                                                                                                                     
   2F68C223A6C     3701 Tue Aug  4 11:30:32  [email protected]                                                                                                                                          
               (TLS is required, but was not offered by host 10.0.0.1[10.0.0.1])                                                                                                                         
                                           [email protected] 4.
   

4. Fix the problem.

5. Rebuild with docker-compose up -d --force-recreate

6. Execute postqueue -p in the container.

Actual Behavior

The new mail queue will be empty.

Running this command seems to explain why:

$ postconf -p queue_directory
queue_directory = /var/spool/postfix

That's the directory where mail queue is stored, and it's not in any volume.

Expected behavior (i.e. solution)

The queue should be stored under /var/mail-state/spool-postfix. Or /var/spool/postfix should be a recommended mount point. Or maybe even /var/spool? Because this command shows that it's used also somewhere else:

$ postconf -p | grep /spool
queue_directory = /var/spool/postfix
smtpd_sasl_path = /var/spool/postfix/private/auth

Your Environment

• Amount of RAM available: 8GiB
• Mailserver version used: If the org.label-schema.vcs-ref label is right, it points to f225e14. In any case, the exact image is sha256:701355354a94a936006df1e7635eec5fff2cedf166c93fb6cffc0663e9ff0e34 downloaded from Docker Hub.
• Docker version used: Docker version 18.06.1-ce, build e68fc7a
• Environment settings relevant to the config: Ubuntu 18.04.1 LTS

@Tecnativa TT25079