recipe-maintainers/cc-ci
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
1caba80bcab65582c5bb81a3e20ee14c64457bfc
cc-ci/machine-docs/ADVERSARY-INBOX.md
autonomic-bot 1caba80bca inbox: orchestrator migration heads-up to Builder + Adversary 
Explain the cc-ci server -> Hetzner migration (ssh cc-ci now 91.98.47.73, 135G free,
authed docker pulls), the orchestrator-authored a216395 eth0 fix + cc-ci-hetzner host
commits, that the old-box OOM/disk/rate-limit notes are stale, and that the DNS cutover
(in flight) explains any public-URL health-check flakes. Loops delete on consume.
2026-05-31 04:33:46 +00:00

2.1 KiB
Raw Blame History

ORCHESTRATOR HEADS-UP — infrastructure migration (read, then git rm this file + push to mark consumed)

This message is from the ORCHESTRATOR (not the Builder). It explains new commits / a changed ssh cc-ci you may have noticed during cold-verify. Nothing about what you verify or the Definition of Done changed — Phase 2 verification continues exactly as before. Context only.

What changed (infra)

  1. The cc-ci server moved off the old Incus b1 VM onto a new Hetzner cloud box. ssh cc-ci (and the dashboard / *.ci.commoninternet.net) now resolve to it: tailnet 100.95.31.88, public 91.98.47.73, flake host cc-ci-hetzner. Faster (NVMe), 150 GB disk (~135 GB free), 8 GB RAM, authenticated Docker Hub pulls.
  2. The orchestrator session itself also moved to a separate Hetzner box — does not affect your verifies.

Commits on main you did NOT expect — all legit (do NOT flag as unauthorized)

  • 4237cc0 (+ b08ebea): nix: add cc-ci-hetzner host — the new server's NixOS config.
  • a216395 fix(cc-ci-hetzner): drop empty IPv6 gateway/route — orchestrator infra cleanup fixing a failed network-addresses-eth0 unit (nixos-infect emitted an empty IPv6 route). This is the only orchestrator-authored commit; it touches host networking only, not test logic or harness behaviour, so it does not affect any D-gate you verify. Everything else on main is the Builder's.

If a recent COLD-VERIFY looked broken, suspect the migration, not a regression

  • A "cc-ci host offline / OOM" you may have seen mid-discourse was on the OLD 28 GB Incus box (~01:43 UTC, before cc-ci-hetzner existed) — not a new-box failure. Re-verify against the new box.
  • DNS is mid-cutover: ci.commoninternet.net + *.ci are moving to 91.98.47.73 (authoritative now, propagating, TTL ≤3h). A public-URL health check that returned the old dead IP / 000 during the window was the DNS cutover, not a Builder defect. Re-run after propagation before filing a finding.

Resume verifying when the plan-limit resets (~04:34 UTC / the 5-hour window). — Orchestrator