Roadmap¶
Junkmesh is an experiment; the roadmap is a list of questions, in the order we intend to answer them.
Phase 0 — Boot (now)¶
- [x] Alpine-based ISO with Yggdrasil + Garage preinstalled
- [x]
junkmesh-setupone-command installer (disk install, key generation, service wiring, firewall) - [x] Documentation site (this site)
- [x] Reproducible ISO builds in CI, published to GitHub Releases
Phase 1 — Cluster¶
- [ ] Three-node reference cluster on real junk hardware
- [ ] Scripted cluster bootstrap (
junkmesh-cluster init/join) - [x] Per-node metrics/status API (
junkmesh-exporter: Prometheus/metrics+ JSON/api/v1/statuson port 3904 — docs) - [x] Mesh-native discovery —
/api/v1/discoveryserves the whole cluster as Prometheus HTTP-SD, derived from Garage membership, so monitoring needs no hand-maintained target list - [ ] Peer-exchange convention so Junkmesh nodes can find each other without relying on public Yggdrasil peers
- [ ] ARM ISO (aarch64) for retired Chromebooks, Raspberry Pis and Mac minis
Phase 2 — Admission without administrators¶
The hard research question: Garage clusters still need a shared rpc_secret
and someone to apply layout changes. Can admission be made collective?
- [ ] Explore quorum-signed layout changes
- [ ] Explore per-cluster admission policies (vouching, proof-of-storage trials for new nodes)
- [ ] Threat-model the open mesh properly
Phase 3 — Feed back¶
- [ ] Write up findings for the Junk Net community
- [ ] Decide: does Junk Net's Brisbane pilot migrate to Yggdrasil, run both, or stay on Nebula?
Note
No dates. Old laptops taught us patience.