- nix: declarative, simple but finicky and unpopular.
- flatcar: all in one sane option but not declarative. Strong emphasis on security, optimized for container workloads.
- tailscale
- rke2 - a modern version of k3s
- Cilium
- Linstor (via Piraeus operator, possibly with K8S api instead of a separate etcd cluster)
- OpenEBS (very simple and a good option for initial bootstrapping, especially for local/hostpath PVs)
- Longhorn: Simple but annoying and slow. The most annoying tihng is that CPU spikes lead to read only FS which fuck things up.
- Clickhouse
- Postgres (with CNPG operator that will cover everything from HA to connection pooling to backups via CRDs)
- React
- Nodejs
- PHP
- OCaml
- Zig
- Go
- Rust (only if you really really really really need C++...)
- Store logs in Clickhouse
- Loki is fine too....
- Vector.dev as collection and delivery pipeline for logs, metrics, traces. (VRL is very performant and robust)
- Prometheus
- Grafana
- Sysdig
- Netdata
- MinIO
- Flux
- ArgoCD
- Terraform