jakman85andClaude Opus 5
a984442cae
Two debt items, both about making a latent problem impossible rather than fixed. === 1. Node timezone pinned to Etc/UTC (node_baseline, `system` tag) === The prod nodes disagreed, measured from k3s's own etcd snapshot annotations: prod-1/2 at -05:00 (Central), prod-3 at -04:00 (Eastern). Functionally harmless — Kubernetes and etcd are UTC internally, and in UTC those snapshots are ~4 minutes apart, which is correct staggering — but it makes cross-node log correlation actively misleading, and this cluster has had two unplanned reboots in two days. `Etc/UTC` chosen on repo convention, not just principle: catalog/plex.yaml records "no other service in the repo sets a timezone ... so use the safe default" with TZ: Etc/UTC, and apps/wave-0/plex.yaml matches (platform/crafty is America/Chicago DELIBERATELY — a user-facing game schedule, not a log clock). Decisive extra fact: THE THREE DEV NODES ARE ALREADY Etc/UTC, so this makes prod match dev rather than inventing a convention. UTC also removes the class — no DST discontinuity, node time == platform time. New knob `node_baseline_timezone`, overridable per environment via group_vars (the same seam as node_baseline_backports_kernel) but deliberately NOT restated there: no environment needs to differ, and a second copy is a second thing to disagree. Guarded by an assert in the role's existing BLOCKED style before the module runs. `--check --diff --tags system -e k3s_hosts=k3s_prod` INDEPENDENTLY CONFIRMS the offset evidence, and is the whole verification — the play was never run for real: k3s-prod-1: -"America/Chicago" +"Etc/UTC" changed k3s-prod-2: -"America/Chicago" +"Etc/UTC" changed k3s-prod-3: -"America/New_York" +"Etc/UTC" changed The same --check against DEV reports the task `ok` on all three — idempotence demonstrated against live hosts, in check mode only. `--list-hosts` confirms `-e k3s_hosts=k3s_prod` selects the three prod nodes (the diagnostic plays take their OWN var; the wrong flag silently targets dev). *** AUTHORED, NOT APPLIED. *** Applying it is a live-cluster action and was deliberately left for the operator to schedule: prod's Longhorn had only just finished resyncing after a node reboot, and changing a node's clock config during storage recovery is not an unattended move. `ansible-inventory --list` was NOT used anywhere — it prints DECRYPTED vault values (the passphrase is mounted into the dev container) and has leaked the prod become-password to a transcript once already. === 2. New L0 check `placeholder-guard-coupling` (R-ENV-02, R-STATE-03) === On 2026-09-05 the PM changed a placeholder in clusters/base/cluster-vars.yaml from `TBD-PROD-S3-BACKUP-ENDPOINT` to `https://TBD-R2-ACCOUNT-ID...`. The guards in bootstrap/standup.sh matched the glob `TBD-*` — a PREFIX. All three silently died, and `resume` would have enabled velero on prod against a nonexistent hostname: installs clean, reports healthy, no backups exist. Found by accident. A guard in one file keyed on the SHAPE of a value in another file, with nothing tying them. The check exercises the REAL predicate — it does not reimplement it. standup.sh is not sourceable (top-level argv parsing hard-exits, cd, export, EXIT trap), so the check extracts the predicate's own source lines and sources only those. Faithfulness is pinned by: discovery of every `*_is_placeholder` (not a hardcoded name), exactly-one-definition, "extracted block is a verbatim substring", a self-containment gate, a canary (must refuse "", must accept a real URL), and a scan for a divergent second copy. If the predicate grows a dependency the check FAILS LOUDLY and names the remedy rather than degrading to testing a copy — the failure mode it exists for. Both sides derive from real files: guarded knobs from a taint trace in standup.sh (`x="$(cvar KEY)"` -> alias hop -> `*_is_placeholder "$x"`); values from base cluster-vars plus each overlay's JSON6902 /data/KEY patches. The independent detector is deliberately NOT the guard: seed idioms UNION the guard's own words with glob anchors STRIPPED — so a prefix-anchored guard cannot pass, and deleting an arm cannot make a value stop being placeholder-shaped. Real tree: 32 assertions, 18 values through the extracted predicate, PASS. NON-VACUITY VERIFIED BY ME, twice, on throwaway copies (the real repo and the read-only spec repo untouched). My first perturbation tripped the self-containment rail instead of the anchoring property, so I redid it shape-preserving — substring globs -> prefix globs, nothing else — and it failed on exactly the right property, naming mid-string and tail acceptances with the diagnosis "This is the 2026-09-05 failure exactly". Control: unperturbed copy exits 0. === 3. ci/touched-requirements.sh: bootstrap/ was UNMAPPED (found by the agent) === The path->requirement map had no `bootstrap/*` entry, so a change touching standup.sh ALONGSIDE any already-mapped path scoped the L0 run to that other path's requirements and silently excluded the new check — the one that exists to catch exactly the 2026-09-05 class. A safety net with a hole where the incident was. Mapped to R-STATE-03 R-GITOPS-01; verified the map now resolves bootstrap paths. VERIFIED: full CI gate PASS. `run.sh --layer l0` 8/8 PASS including the new check. Coverage unchanged at 45/51 (the new check tags existing requirements). Refs R-ENV-02, R-STATE-03, R-TEST-01, R-PLAT-01. Co-Authored-By: Claude Opus 5 <[email protected]>
home-network-k3s-impl
Implementation of the home k3s + Longhorn platform. This repo holds the reconciled artifacts
(Flux config, Kustomize overlays, KubeVirt/Helm manifests, ServiceDefinitions, Kyverno policies,
Chainsaw conformance tests, CI). It is driven by — and subordinate to — the read-only
specification at ../home-network-k3s.
Rules (from the spec — non-negotiable)
- The spec is read-only. Never modify anything under
../home-network-k3s. Implement against it; escalate spec problems, don't edit them (specAGENTS.md§11, R-TEST-06). - Done = a green, reproducible, requirement-tagged conformance test — never "merged," never a
manual demo (spec
docs/14,agents/definition-of-done.md). - No work on
main. Every change is a feature branch merged intomain(specCONTRIBUTING.md). - No plaintext secrets — SOPS-encrypted only (spec
docs/07). - Phase discipline — only build what the active roadmap phase authorizes (spec
roadmap/phases.md).
Layout
infra/ node provisioning: OpenTofu (Proxmox VE) + cloud-init + Ansible, and the
pinned dev container that carries the whole toolchain
bootstrap/ one-time cluster bootstrap: Flux install, SOPS/age key, host Gitea
platform/ platform CONTROLLERS (Flux HelmReleases / operator bases): kyverno, cilium,
longhorn, external-snapshotter, cert-manager, externaldns, gateway, coredns,
vpn-egress, monitoring, cnpg, keycloak, oauth2-proxy, kubevirt
platform-config/ platform CONFIGS — the custom resources of the CRDs installed above
(ClusterIssuers, Gateway, Kyverno policies, KubeVirt/CDI CRs, Keycloak CRs).
Split out because CRs and the CRDs they use cannot apply in one pass.
apps/ application tier, reconciled as dependency-ordered WAVES:
wave-0/ wave-1/ wave-2/ rendered container apps (see DESIGN-waves.md)
vms/ KubeVirt VirtualMachines (non-gating tier)
catalog/ ServiceDefinition entries (the onboarding surface) + render.py
schemas/ ServiceDefinition JSON Schema + renderer contract
clusters/ Flux entrypoints + dev/ and prod/ overlays
conformance/ Chainsaw behavior/conformance suite + harness + coverage report
ci/ the gate as scripts (lint · schema · Kyverno L0 · L1 render · decision)
.gitea/ Gitea Actions CI running that gate on an in-cluster runner
Status
Roadmap phases 0–6 are delivered and deployed on the dev cluster; Phase 7 (prod) is not
open — clusters/prod is a design-only overlay.
| Phase | Delivered |
|---|---|
| 0 Foundations | ServiceDefinition schema + renderer, Kyverno L0 gate, conformance harness, CI |
| 1 Cluster bring-up | 3-node k3s HA on Proxmox VE (192.168.1.50/.51/.52), Cilium, Flux, SOPS |
| 2 Platform core | Cilium · Longhorn · cert-manager (LE-prod wildcard) · ExternalDNS · Gateway API · split-horizon CoreDNS |
| 3 Egress / VPN | ADR-0015 node-level OpenVPN split-tunnel + Cilium egress-gateway steering + host kill-switch |
| 4 Observability | kube-prometheus-stack · Loki · Alloy · Alertmanager. Velero still blocked on OQ-BKP; image-automation authored but disabled |
| 5 IaaS | KubeVirt + CDI on real nested KVM; live-migratable RWX VM storage. The UniFi VM runs, but UniFi OS Server is not yet installed in the guest |
| 6 Identity + CI | Keycloak (Operator + CNPG) · oauth2-proxy forward-auth · Grafana OIDC · kubectl OIDC SSO with group→RBAC · in-cluster Gitea Actions merge gate |
Conformance: 36 / 51 requirements covered (70.6%), 15 gaps, 1 safety-critical gap
(R-STATE-02, Longhorn node-loss). Regenerate with ./conformance/run.sh --coverage-only;
the full report is conformance/_artifacts/coverage/coverage.md.
The running narrative — decisions, root causes, incidents, follow-ups — is
DECISIONS-LOG.md. It, not this table, is the authoritative record.
Working on this repo
Everything the gate needs is baked into the pinned dev container — do not install the toolchain on the host:
infra/devcontainer/dev.sh build # once
infra/devcontainer/dev.sh run make -C ci gate # the same gate CI runs
infra/devcontainer/dev.sh # interactive shell (sops, flux, kubectl, tofu, ansible)
kubectl/flux/helm in there talk to the dev cluster by default. Prefix CLUSTER=prod to
target prod instead:
infra/devcontainer/dev.sh run kubectl get nodes # dev (default)
CLUSTER=prod infra/devcontainer/dev.sh run kubectl get nodes # prod
Both clusters are always present in the container's KUBECONFIG as the contexts k3s-dev and
k3s-prod; CLUSTER picks which one is current, and each run prints the cluster it is targeting.
The container mounts the repo at /work, your age keyring, and the sibling spec repo read-only
so the spec-immutability check can actually run. See
infra/devcontainer/README.md.
A green run prints MERGE GATE: PASS. The L2/L3 conformance layers stay SKIPPED unless
CLUSTER_AVAILABLE=true — the static gate never pretends to have run them.