Running this Blog on Nomad

In an attempt to consolidate my various personal projects, and to be efficient about how much money I spend on VPS hosting, this blog is now running on my tiny Nomad cluster. The ID for this allocation is:

The components involved are:

  1. Nomad
  2. Consul
  3. Vault
  4. Fabio
  5. multirootca

The Nomad deployment guide recommends either three or five servers, but I’m not really running business-critical applications, so I currently only have one server and one client node.

The server node is running one instance each of Consul, Nomad, and Vault, the first two in server mode, with certificate authorities defined on a central support server.

I save all of the config files in my infrastructure repo. In particular, these job files are responsible for running this blog:

  1. damienradtkecom.nomad (Hugo server)
  2. fabio.nomad (load balancer)
  3. acme-renewer.nomad (certificate renewer periodic batch job)


This job is responsible for running hugo server on the blog’s source directory. It specifies the service tag expected by Fabio so that requests to get routed to the blog server.

It also runs two instances and specifies an update block to ensure zero-downtime deployments.


This job runs the Fabio load balancer on a randomly-assigned port so that it doesn’t require root privileges, along with a custom, tiny Go program running as root that routes traffic from port 443 to Fabio.

One upside to having only one client node is that the domain’s A record can be set to the client node’s IP address, so traffic will properly make its way to fabio. In case of a multi-client cluster, one node will need to be designated the load balancer node, and the fabio job configured to always run on it.


This is a periodic batch job that uses to renew the domain’s SSL certificate using a DNS challenge and the Linode API. The results are stored in Vault’s KV store, which Fabio is configured to read from to support HTTPS.