Recap of tech choices: Cilium networking, MetalLB load balancing, nginx ingress, Rook/Ceph storage, Prometheus/Grafana monitoring/alerting, Loki logs, Harbor container registry, Flux for GitOps.
Also a huge shout out to the Ingress controller comparison spreadsheet[1] the author made! Really nice being able to compare these feature sets. Neat to see a couple folks already working on HTTP3 & QUIC load balancing, who does Maglev routing, who has the best offload capabilities. This definitely brought my interest in Apache APISIX way way up.
The author digs at their own infrastructure some, calling it for learning. But to me, I think it's bad/dangerous having a conservative outlook that personal users should have to learn/use inferior/lesser tools (dozens of hours), it's bad to bifurcate the computing world into high and low computing. It took the author a lot of effort I'm sure to cobble together this system, but my hope is, over time, we start to make works like this much more paved road, and we start to make them much more accessible & documented & supported. We still haven't seen personal-use-focused Kubernetes distributions arrive, but I'm hoping that happens. Here's the author's description:
> By trying out "flashy" software in a low-stakes environment, I can have some idea of how it will work in production-critical ones, which helps me make better choices. (As an example, my attempts to run Istio in a homelab setting have convinced me that it should be avoided in most circumstances.)
Again I think this is undershooting the real value, of using real tools. The hill to climb right now seems big. But imo the effort one's going to invest in selfhosting is probably going to be significant, and I like a lot the plan of shooting for good.
[1] https://docs.google.com/spreadsheets/d/191WWNpjJ2za6-nbG4ZoU...