Deploying Multiple ZoneMTA Containers on single server

[muzammalvoltic]

Hi, I successfully converted the ZoneMTA application to a Docker container, and the single container works fine. However, I am now trying to deploy multiple ZoneMTA containers. Could you please guide me on how to achieve this?

I tried the following approach:

Used HAProxy and Traefik Proxy to bind the same port to multiple containers.
However, while the containers run, they never receive or send any emails.

Any assistance on how to properly configure multiple ZoneMTA containers would be greatly appreciated.

[NickOvt]

There could be multiple possible approaches

1. Replicas
2. Separate containers.
   It all boils down to what exactly you want to achieve?
   If you want to have multiple zone-mta servers with same port and have a loadbalancer then you'd have to have traefik setup for example, working as a load balancer

[muzammalvoltic]

There could be multiple possible approaches

1. Replicas
2. Separate containers.
   It all boils down to what exactly you want to achieve?
   If you want to have multiple zone-mta servers with same port and have a loadbalancer then you'd have to have traefik setup for example, working as a load balancer

I initially created the ZoneMTA container, but it only works when I use host network mode, like this: docker run -d --network host zonemta. Mapping just port 587 doesn’t seem to work. As for Traefik, I tried it, but it didn’t help. Both containers start, but they can’t fetch or send emails. I believe this is because the ZoneMTA container needs host network access to fetch IPs for sending emails.

In short, I aim to set up separate servers for ZoneMTA, deploying multiple containers, each with its own set of IPs. However, based on my current testing, this approach hasn’t worked as expected.

[ErgjanJaha]

@muzammalvoltic that is something i have actually solved quite fine.
You can start by completely splitting the feeder (receiver) containers from the senders. You can smtpInterfaces.feeder.enabled: true, remove all the zones and set queueServer.enabled: false
You can run the feeder anywhere without having to worry about ip addresses ( I even run it on Kubernetes )

Then you can do the opposite of that in the servers where you have configured your Outbound IP Addresses.
Disable the feeder, you can use default zone ( if you are only using a single server ) / otherwise multi custom zone names and InstanceIDs is a requirement.

This means you can run the instances which receive the emails and queue them, and the workers that send those emails from the queue, completely separately. You can scale them independently and so on.

You also don't need to have one ip per container, you can just assign all ips on all containers and just increase the number of replicas.

[muzammalvoltic]

@ErgjanJaha Thanks! This was pretty complex for me, but I managed to figure it out. Here's how I solved the issue:

My Solution:

1. Set up multiple ZoneMTA containers - I deployed several instances instead of just one
2. Use host network mode - Applied network_mode: host to all containers for direct IP access
3. Distribute IPs across containers - I assigned 4 IPs to each container (adjust the API ports and other settings to avoid conflicts)
4. Share the same database - Connected all containers to a single database so they all process the email queue together

Important Notes:

• Make sure to change API ports and other configuration settings for each container, otherwise they'll conflict
• I couldn't control which specific email gets sent from which zone/container, but that wasn't my main goal anyway

What I Achieved:

My goal was to deploy ZoneMTA across multiple servers (or multiple instances on a single server) to:

• Increase the total number of available IPs
• Distribute email processing load across multiple instances
• Scale horizontally for better performance

This setup works great for my use case!
And yeah if I'm doing something wrong feel free to guide me, Thanks again.