Free, private pi-hole hosting with and Tailscale

Hosting a network-wide pi-hole for ad-blocking is easy. It comes with a one-step installation guide that you can run on most environments. I run two pi-holes at my house, but the problem with my setup is, if my internet drops, both pi-holes stop working too. Ideally, I’d need to have a pi-hole outside my house, preferably where internet doesn’t go down at the same time as my provider.

That’s where a cloud-hosted pi-hole proves helpful.

I learned about recently. It’s an app hosting platform that makes it incredibly easy to deploy apps to multiple regions and scales as needed. Their free tier offers a generous 2,340 hours per month of uptime, which translates to about 3 shared-cpu-1x VMs with 256MB RAM full time. My pi-hole consumes about 200MB RAM consistently, so this means I could host upto 3 apps each with pi-hole on it. But I have only one as I have two others at my house.

A screenshot from's pricing page that shows their free tier metrics. free tier

Setting up the pi-hole

This Fly blog post already documents how you can run a pi-hole in a few steps, but there is a problem with this setup: the pi-hole will be publicly query-able by anyone on the internet. We don’t want that because a public, open DNS resolver is not a good idea. We’ll need to lock this down using a secure tunnel that’s accessible only to you.

Locking down pi-hole access with Tailscale

If you are following my blog, you’d know by now that I am a Tailscale fan. It’s an easy, useful mesh VPN software that you can add to most devices that you have. For the pi-hole on setup, I followed the same guide as Fly documented, but in the Dockerfile configuration, I replaced eth0 with tailscale0 so that my pi-hole listens for queries only on the Tailscale network.

With that image deployed, I SSH’d into the Fly instance and installed Tailscale using the Debian installation guide here: Install Tailscale on Linux. That’s not quite easy though. I had to use legacy iptables and then run ./tailscaled & under usr/sbin folder. From there, I could run sudo tailscale up.

Stopping ./tailscaled stops Tailscale, so as a workaround for now, I just close the tab where ./tailscaled is running. I am pretty sure that’s not how I must be doing it, but it works for now.

I further locked down this Fly instance’s DNS port access to my Tailscale nodes as I don’t want anyone else on my Tailnet (I share it with my friends and family) to make other calls to the Fly node. A handy Tailscale ACL like the one below works:

{ "Action": "accept", "Users": ["group:not-arun-family"], "Ports": ["fly:53"] },

Leave a reply