Closed geerlingguy closed 4 years ago
(Just noting that a PV is required for persistence, and the K3s cluster includes rancher.io/local-path
, so as with the MySQL databases, a reboot could really throw things for a loop if Pods that require local PVs are scheduled on different nodes than where they originally started!)
So... Pi-hole is slightly heavyweight as a K8s-style app, it's better to run it directly using Docker and/or standalone. Though I had it working. The Helm chart for it (linked previously) is fairly complete but was not as simple as I was originally hoping, and I want it to be simple and fast.
I'm going to look into a newer/slimmer alternative, blocky
, which even has prometheus metrics so stats can be aggregated in a Grafana dashboard! https://github.com/0xERR0R/blocky
Though I might go back to pihole...
Blocky was about the same experience as Pi-hole. Slightly lighter weight and faster to start... but I do enjoy the UX of Pi-hole a bit more.
Woohoo! Using the following values, I can connect via a LoadBalancer
in the cluster:
serviceUDP:
loadBalancerIP: '10.0.100.99'
type: LoadBalancer
And I can add any of the worker node Pis to my DNS Servers list on my Mac to connect through Pi-hole:
And logged in as admin:
Definitely sticking with Pi-hole for the fancy UX.
Testing with dig
shows 28-32ms query time for sites like apple.com/google.com.
Using my default router / ISP DNS, I get 30-48ms... maybe my ISP (Spectrum) ain't so good after all! I sometimes switch to 8.8.8.8/8.8.4.4, but for now I'm going to stick with the Turing Pi Pi-hole instance a while and see how it works out.
As another example of a nice little utility that can be run on a small Pi cluster (but strikes way above it's weight class in effectiveness and utility), I'd like to install Pi-hole on the cluster and see how it runs.
I'm thinking of just using this Pi-hole chart: https://github.com/MoJo2600/pihole-kubernetes/tree/master/charts/pihole