Operator decision on PR #81: no CI job may reach a privileged daemon. The act runner is untouched.
- platform/buildkit (+ clusters/base/buildkit.yaml): moby/buildkit v0.34.1-rootless pinned by
digest, uid 1000, not privileged; caps drop ALL + SETUID/SETGID, seccomp/AppArmor Unconfined,
read-only rootfs (each measured under podman). cert-manager CA + server/client certs, ClusterIP
:1234, default-deny with ingress only from the act-runner pod (DinD masquerades jobs to it).
No Kyverno exception: every applied policy matches app-namespace=true only.
- Each Containerfile names its shipped stage and ends in a `verify` stage holding every former
`docker run` assertion (uid 65532, RUN --network=none, writable-path inventory standing in for
the read-only rootfs). The old `&& echo` checks that could never fail are real assertions now.
- images.sh drives buildkitd with a pinned buildx (sha256-checked) via bake; PRs build verify
targets with agent-runner FROM the in-run toolbox (target: context); main verifies, then pushes.
Fails loudly when the builder secrets are absent or buildkitd is unreachable.
- infra/gitea/buildkit-client-secret.sh copies the client cert into Actions secrets.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Operator rule 2026-10-29: image creation, hosting and lifecycle belong to CI/CD, never to a
workstation build/push. Nothing had ever been pushed to the Gitea registry.
- .gitea/workflows/images.yaml + scripts/images.sh: PRs build+verify (no credential); main
builds, verifies and pushes toolbox, agent-runner (FROM toolbox) and conformance-runner as
sha-<commit> + main, reading the digest back from the registry. Only changed images build.
- Each image's own build-image.sh does the build and in-image verification, now engine-neutral
(DinD: no podman-only calls, probe over stdin not a bind mount). New toolbox build-image.sh.
- Digest references carry `# renovate-digest:` markers; renovate.json5 gains a custom.regex
manager for them. The worker Job template is deliberately untracked (it would loop).
- renovate.yaml pins renovate-version 44.167.0: the action's default (39.264.0) rejects the
current config's managerFilePatterns, so every run would have died at CONFIG_VALIDATION.
- Push credential: REGISTRY_PUSH_TOKEN secret; the Actions token has no package scope in
Gitea 1.24.3 (container.go Authenticate). Retention: infra/gitea/registry-retention.sh.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Operator correction: "sop, wg, and any other tool requirement is supposed to be
baked into the devcontainer."
My previous commit discovered that `wg` existed only on the host, `sops` only in
the container, and `qrencode` on neither -- and fixed it the wrong way, by
splitting every step of the key procedure across the boundary and telling the
operator to `dnf install qrencode` on their workstation. That breaks CLAUDE.md §5
and the devcontainer README's "Don't install the toolchain on the host", produces
a procedure that cannot be reproduced in CI, and is miserable to follow.
wireguard-tools + qrencode are now distro packages in the Containerfile. Both are
in Fedora 44's repos (wireguard-tools 1.0.20260223 -- the same upstream version as
the image this component runs -- and qrencode 4.1.1), unpinned like every other
dnf package there: the pinned thing is the Fedora base, and neither is a gate tool
whose version changes what green means.
Three things got SIMPLER rather than merely relocated, which is the tell that the
rule was right:
* Step 2 is one pipeline in one shell -- `sops set ... "$(wg genkey)"` writes the
key straight into the already-encrypted document. No staged copy in .cache, no
`*.key`-suffix warning, no sha256 comparison to prove the right bytes landed.
* new-peer.sh derives the server public key from wg-server-key.sops.yaml itself
(`sops -d --extract | wg pubkey`), so there is no local key copy to keep in
step with the encrypted one, and no magic host path it must live at.
* `.cache/` is gitignored at the repo root (mirroring infra/ansible/.gitignore),
giving the container a persistent non-committable scratch dir. new-peer.sh
writes profiles to .cache/wg-peers/ -- necessary because the repo bind-mount is
the container's only persistent read-write mount, so a profile in /tmp there is
gone before it can reach a desktop. Its blanket "refuse to write inside the
checkout" guard is now a `git check-ignore` test of the real property.
Verified by execution, not reasoning: the image rebuilds; wg/qrencode/sops all
resolve in it; `sops -d | wg pubkey` runs in one shell; new-peer.sh refuses the
REPLACE-ME placeholder with an actionable message; and a full run with a throwaway
server key renders a correct profile at 0600 under a 0700 gitignored dir, prints
the QR, and refuses a non-ignored destination inside the repo.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
The CNPG alert unit tests (324 lines, mutation-checked) were run by NOTHING. promtool was
not in the baked toolchain, no gate stage evaluated any *.promtool-test.yaml, and the only
runner `curl | tar -xz`'d a promtool release from GitHub at TEST-RUN time if one was
missing. Hours later a second, independent harness landed for the Velero rules
(platform/monitoring/tests/) with its own fallback — a `podman run` of
quay.io/prometheus/prometheus. So the repo had two harnesses, two acquisition paths, and
coverage that looked present and was enforced nowhere: the exact failure mode that had just
produced a permanently-firing CNPG alert and a structurally inert Velero alert.
PROMTOOL IS NOW PINNED LIKE EVERY OTHER TOOL. PROMTOOL_VERSION=3.14.0 in
infra/devcontainer/versions.env and ci/versions.env (the two stay in lockstep, per the
repo's rule), baked by the Containerfile from the prometheus release tarball (promtool ships
inside it), recorded in /etc/dev-toolchain-versions, passed by dev.sh, and installed for the
l0 stage by .gitea/workflows/scripts/install-tools.sh so the CI runner is not permanently
SKIPPED. The pin is not free-floating: promtool IS the rule engine, so it tracks the
Prometheus prod actually runs — verified against the live cluster, not assumed:
kubectl --context k3s-prod get prometheus -n monitoring -o jsonpath='{.items[*].spec.version}'
-> v3.14.0-distroless (image quay.io/prometheus/prometheus:v3.14.0-distroless)
ONE HARNESS. platform/cnpg/run-alert-tests.sh is folded into
platform/monitoring/tests/run.sh, which now covers every PrometheusRule in the repo that has
tests, via two conventions it discovers automatically: `<manifest>.promtool-test.yaml`
beside its manifest (cnpg), and `<component>/tests/*.test.yaml` over that component's
manifests (monitoring). It writes the flattened rules under whatever name each test file's
own `rule_files:` declares, so no test file needs to know the harness exists, and it
`promtool check rules` first — a rules file that does not parse would make every unit test
pass vacuously by finding no rules at all. The cnpg script's reasoning is preserved: why a
CR must be flattened (`.spec.groups`, never re-typed), and why generated files go to a temp
dir. Its `${BASE_DOMAIN}` substitution is dropped in favour of extract_rules.py's verbatim
placeholders, with the trade-off recorded there — the hazard it guarded against is addressed
by the temp dir, and verbatim leaves one transformation fewer between the shipped rule and
the rule under test.
NO ENGINE IS ACQUIRED AT RUN TIME any more — not the GitHub release, not the container
image. Not because a pinned download is unusual here (the Containerfile downloads every tool
it bakes) but because those two decided the version, the network and the machine SILENTLY,
at test time, on whatever host ran the tests; and each quietly turned "the toolchain has no
promtool" into a green run, which is how the toolchain stayed without one. Missing promtool
is exit 3 (UNWIRED, never green); a PROMTOOL= that does not resolve is exit 2.
ONE WIRING: conformance/tests/l0-static/alert-rules-unit-tested, NOT a gate stage. The
`prom-rules` stage (ci/prom-rules.sh + its Makefile target + ci/gate.sh entry) is deleted:
it is the bespoke-per-check pattern b9f8807 had just removed from ci/l0.sh, and the
declarative replacement gives what a hand-written stage cannot — selection by
`conformance.home.k3s/scope: "whole-repo"` (asserted: the test is selected even under a
requirement filter matching nothing), requirement-attributed JUnit/coverage/ledger evidence,
and run_l1_offline.py's rule that a SKIP of a scope-declaring check is the runner's rc 3.
The scope declaration is load-bearing here: touched-requirements.sh keys apps/* to nothing
and platform/monitoring/* to R-OBS-01/02, never R-OBS-04, so a new untested PrometheusRule
would otherwise be gated by nothing.
The check asserts the pins and their lockstep, that the image bakes and records promtool,
that the promtool IN USE reports exactly the pin (a FAILURE, not a skip: the wrong engine
still prints SUCCESS), that exactly ONE harness exists and none fetches an engine, that
every rule-unit test file passes, and — on every run — that the harness CAN fail, by putting
a synthetic rule through it twice with one expectation mutated.
MISSING TESTS ARE A RATCHET, deliberately, and the reasoning is in the check's header. 51 of
58 alert rules have no unit test. Failing on all of them makes the gate red until ~51 tests
exist, so it gets disabled and protects nothing; merely reporting is exactly how the CNPG
tests came to be enforced by nothing. So the gap is DECLARED in the check as a shrink-only
register, printed with counts every run (13 alert files nothing tests at all; 35 named
alerts inside covered files that nothing asserts on), and it cannot grow: a new
PrometheusRule, or a new alert in an already-covered file, FAILS without a test, and a stale
entry FAILS too. A missing test is never an exit-3 skip — SKIPPED is reserved for the one
genuine environment gap, promtool being absent.
PROVEN, in the rebuilt image (promtool 3.14.0):
* both suites pass through the single harness: cnpg 6 rules, monitoring 36 rules, SUCCESS.
* the check passes: 95 assertions, 7/58 alert rules unit-tested, 13 declared gap files.
* NOT VACUOUS, twice over. Shortening CNPGPrimariesCoLocated's `for: 30m` to 5m turns it
red (pending expected at 25m/29m, got firing), reverted. Adding an untested
PrometheusRule (platform/kubevirt/probe-alerts.yaml) turns it red on the ratchet,
removed. The harness's own mutation self-test runs on every invocation.
* the scope mechanism selects it: `run.sh --layer l0 --requirements r-nosuchreq-99
--include-scope whole-repo --print-selection` lists alert-rules-unit-tested.
Co-Authored-By: Claude Opus 5 <[email protected]>
Every unprefixed './dev.sh run kubectl ...' pointed at the dev cluster,
which was decommissioned on 2026-09-12. The banner cheerfully named dev and
the command then hung or errored against a context that no longer answers,
with nothing saying the cluster is gone. I hit this repeatedly tonight.
CLUSTER=dev is still accepted rather than removed, because the dev VM disks
were deliberately kept so the environment is reversible — but it now warns.
Co-Authored-By: Claude Opus 5 <[email protected]>
infra/devcontainer/kubeconfig.sh was never staged into the delivery-A commit, so
the branch built on main carried main's single-cluster tooling and `--context
k3s-prod` failed with 'context does not exist' mid-incident. Not Flux-reconciled,
so it never affected the cluster — but it breaks prod diagnosis from a checkout of
this branch, which is when you need it most.
Carries the concurrency fix: the merged kubeconfig holds both contexts and NO
current-context, with each invocation getting a per-invocation selector file
prepended to KUBECONFIG. Two concurrent dev.sh runs with different CLUSTER values
previously overwrote each other's current-context mid-command — reproduced live,
a CLUSTER=dev reader returned k3s-prod nodes on 3 of 7 iterations.
Co-Authored-By: Claude Opus 5 <[email protected]>
Text-only cleanup after the upgrade campaign. Updates stale version
strings, prose, and comments to the verified current facts; no
Kubernetes behavior or version pins changed except the kubectl
toolchain bump (below).
Verified facts applied: k3s v1.36.4+k3s1, Debian 13 trixie,
Cilium 1.20.1 / Gateway API v1.6.1, Longhorn 1.12.1, node RAM 24 GiB,
dedicated virtio2 SSD etcd disk per node, Velero chart 12.x / v1.18,
Grafana 13.2.0.
- Version strings across infra/ and platform/ READMEs + comments
- kubectl toolchain 1.31.4 -> 1.36.4 (devcontainer/ci/conformance
versions.env; 1.31 vs a 1.36 cluster exceeded the +-1 skew)
- Retired-single-apps-Kustomization refs -> apps-wave-0/1/2 + apps-vms
with per-wave SOPS decryption (flux/clusters docs + bootstrap.sh)
- COMPLETED/historical banners on the three executed plan docs
- Resolved TODO(WS-SEC) markers -> Renovate-maintained ranges
- Appended dated DECISIONS-LOG entry (2026-08-30) superseding the
2026-08-26 LIVE STATE line
Left as revalidation items (not blind-edited): external-snapshotter
v8.2.0 vs 1.36, Cilium ext_authz/forward-auth relevance under 1.20.1,
conformance ImageVolume feature-gate assertions under 1.36.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
`ci/touched-requirements.sh` documents its detection as "R-* tokens literally
present in the diff of changed files". It did not do that — it grepped the ENTIRE
CONTENT of every changed file. So a change inherited every requirement those files
had ever mentioned in prose, and `--require-all` then blocked the merge on
requirements the author never went near.
On this branch that meant editing the decision log and the tofu docs dragged in
five unrelated requirements, all on untouched lines, all legitimately uncovered:
[coverage] REQUIRE: FAILED — required requirement(s) uncovered: [5 IDs]
The practical effect was that the PM could not append to the decision log without
failing CI — which is a fair part of why the log went stale in the first place.
This is the defect already logged as "FOLLOW-UP (WS-TEST): fix the matcher so real
requirements need not be exempted".
Two scoping rules, both learned from the gate blocking merges it should not have:
1. Scan the ADDED/REMOVED lines of the diff, as the header always claimed.
Removed lines count too — deleting a test's requirement tag, or a policy's
self-declared ID, touches that requirement just as much as adding one.
2. Skip `*.md`. An ID in prose is a REFERENCE, not a declaration: explaining a
requirement is not changing behaviour against it. Rule 1 alone is not enough —
a commit that documents a mis-attribution names the IDs involved and thereby
re-triggers the very mis-attribution it fixes, and the decision log can never
record which requirements a change concerned.
This TIGHTENS attribution rather than weakening the gate:
- no requirement is exempted, and DOCUMENTED_GAPS is untouched;
- authoritative declarations (`conformance.home.k3s/R-*` tags, Kyverno policies,
ServiceDefinitions) all live in YAML, on lines a real change must touch;
- the coarse path->requirement fallback is unchanged and file-type agnostic, so
the doc subtrees that matter (conformance/, schemas/, catalog/, clusters/, ci/)
stay covered either way.
Alternatives rejected: adding the five IDs to DOCUMENTED_GAPS genuinely weakens the
gate; scrubbing requirement IDs out of prose (precedent f9e9640) would mean editing
the decision log's own history to appease a matcher.
Gate semantics change — flagged in DECISIONS-LOG.md for review.
Also drops three R-TEST-06 mentions this branch had added to prose and refreshes
the now-stale "gate is RED" notes.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
The docs had drifted badly behind ~85 commits of delivery. The root README still
said "Phase 0 — Foundations. No cluster exists yet"; the infra docs still
described a libvirt/KVM cluster that has been retired for two days; and
DECISIONS-LOG.md — the PM's authoritative record — stopped at 2026-08-25 AM.
Anyone following these docs would have been working from a false map.
Verified against the live cluster (3/3 nodes, 7/7 Flux Kustomizations Ready,
10/10 HelmReleases Ready) rather than inferred from commit messages.
README.md
- Phase 0 -> a per-phase status table: 0-6 delivered and deployed, 7 not open.
- Layout now shows the wave split, platform-config, apps/vms, infra, bootstrap.
- Real conformance number (36/51, 70.6%) and how to regenerate it.
- Devcontainer workflow; note that the gate was red and why.
infra/PHASE1-STATUS.md — rewritten. Was a `virsh`/libvirt resume guide for a
cluster that no longer exists. Now: the Proxmox substrate, the pfSense NAT
paths, both kubectl access paths (Keycloak OIDC normal / cert break-glass),
the toolchain, off-repo state that must not be lost, and the fact that Flux's
git source is an app inside the cluster (`git push origin main` IS the deploy).
infra/PHASE1-SCOPE.md — historical banner. The exit criteria and sizing
rationale are still worth keeping; the libvirt execution steps are not.
infra/README.md — Proxmox, plus a map of the subtree and where the cluster
actually lives (it is not on this workstation).
infra/devcontainer/README.md — the "run tofu and ansible on the HOST" limitation
is OBSOLETE: it was a dmacvicar/libvirt polkit/PID problem, and bpg/proxmox
talks to an HTTP API. Everything runs in the container now. Documents the full
CI-gate toolchain and what each bind mount is load-bearing for (the age keyring
and the read-only spec repo especially — without the latter, spec-immutability
silently skips instead of verifying).
infra/tofu/README.md — was "Authoring only — nothing here has been applied". It
is applied and live. Disk sizes corrected 40/100 -> 64/128 GiB (confirmed
against the running nodes), pfSense NAT marked done, and a warning that
terraform.tfstate is absent on a fresh workstation — an apply with empty state
would try to recreate the three live nodes.
infra/tofu/variables.tf — section comment disagreed with its own defaults.
infra/cluster.sh — retired-with-instructions header. It silently drives libvirt
domains that do not exist; behaviour unchanged, but it no longer lies.
apps/vms/unifi/README.md — was "NOT WIRED IN", 60Gi, and listed a file that has
since been dropped. Now: live status, 40Gi, resolved open questions marked
resolved, and the two constraints the next person will otherwise rediscover —
KubeVirt's 2048B inline userData cap, and that apps-vms must never carry
postBuild.substituteFrom (envsubst corrupts the cloud-init).
DECISIONS-LOG.md — backfilled the gap as one structured entry: the Proxmox
migration, NAT + kube-vip, the twice-proven oauth2-proxy cold-start deadlock,
the Flux source cutover, CI going live in-cluster, ten app onboardings and the
renderer bugs they exposed, the wave split (verified: DBs adopted, not
recreated — original ages intact, prune restored), KubeVirt storage, the UniFi
VM, and kubectl SSO. Plus verified live state and the carried-forward open list.
Two corrections to the record while verifying:
- kubectl OIDC is LIVE (`kubectl auth whoami` -> oidc:<user>,
[oidc:platform-admins]), so the rolling restart 3472f1d deferred has
happened.
- UniFi OS Server is NOT installed in the guest. The VM is healthy (Ubuntu
24.04.4, guest agent connected, live-migratable) but :11443 answers nothing,
and the CNP explicitly permits cluster -> :11443, so it is genuinely not
listening rather than a policy artifact.
Docs and comments only; no manifest semantics change.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Controller now reaches the kube-API directly through the pfSense NAT
(192.168.2.50:6443 -> kube-vip VIP 192.168.1.49:6443), validating TLS
against the VIP via tls-server-name. Removes the dev.sh SSH-tunnel
scaffolding and updates the ansible source-of-truth so a rebuild
regenerates a direct-access kubeconfig.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Bake the conformance harness toolchain into infra/devcontainer so `make -C ci gate`
runs identically here and in CI:
- Containerfile: install kyverno CLI (1.13.4), kyverno-json, chainsaw (0.2.12),
yamllint (1.35.1), check-jsonschema (0.29.4), yq (4.44.6) via new pinned RUN block;
add diffutils + python3-pip; drop stale LIBVIRT_DEFAULT_URI; refresh header for the
Proxmox(bpg)+conformance toolbox; extend /etc/dev-toolchain-versions provenance.
- versions.env: add KYVERNO/KYVERNO_JSON/CHAINSAW/YAMLLINT/YQ/CHECK_JSONSCHEMA;
align KUBECTL 1.32.0 -> 1.31.4. dev.sh build(): pass the new --build-args.
- ci/versions.env: pin CHECK_JSONSCHEMA_VERSION=0.29.4 (single source of truth).
Fix harness seams so the gate can run:
- dev.sh run(): mount the sibling spec repo (home-network-k3s) read-only at
/home-network-k3s for ci/spec-immutability.sh (R-TEST-06); conditional so the
container still runs without it.
- conformance/lib/common.sh conf_dirs_for_layer(): accept a comma/space list of
layers and union the dirs, so ci/conformance.sh `run.sh --layer l2,l3` selects
both layers in one pass (backward compatible for single layers).
NOTE: ci/versions.env pins KYVERNO_JSON_VERSION=0.0.5, which has no upstream release
(max tag is v0.0.3; platform/kyverno/scripts/run-l0.sh already defaults to v0.0.3).
Baked v0.0.3 and flagged for PM to correct ci/versions.env. kyverno-json is not used
by any static gate stage.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
- drop stale libvirt socket mount + LIBVIRT_DEFAULT_URI (old KVM cluster)
- auto-open an SSH tunnel (local :16443 -> primary 127.0.0.1:6443 over the
pfSense SSH NAT) on each run so kubectl/flux/helm reach the cluster with no
manual setup; DEV_NO_TUNNEL=1 opts out, K3S_SSH_* env overrides the NAT front
- refresh header/usage comments (bpg/proxmox, not dmacvicar/libvirt)
Co-Authored-By: Claude Opus 4.8 <[email protected]>
- inventory routes SSH via 192.168.2.102:2250/2251/2252; new k3s_node_ip carries
each node's real 192.168.1.x for --node-ip / k3s_api_host / Cilium parity
- dev.sh bind-mounts ~/.config/home-k3s/ansible-vault-pass (ro) so vaulted
playbooks can run in the pinned container (ansible.cfg vault_password_file)
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Two dev.sh gaps surfaced while running bootstrap/sops/install-age-key.sh:
- kubectl defaulted to localhost:8080 (connection refused) because KUBECONFIG
was never set in the toolbox. Set KUBECONFIG=/work/infra/ansible/.cache/dev.kubeconfig
(override via DEV_KUBECONFIG); --network host already reaches 10.73.0.11:6443.
- age-keygen.sh writes the PRIVATE key to ~/.config/sops/age/keys.txt, but that
path was not mounted in the --rm container, so the key was lost on exit. Mount
${HOME}/.config/sops -> /root/.config/sops (created first) so the key persists
on the host and SOPS_AGE_KEY_FILE resolves to it.
Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
Cluster is UP (3/3 Ready, Cilium OK); Phase 1 paused before Flux bootstrap
(Option A: suspend platform/apps until Phase 2). Add infra/PHASE1-STATUS.md
(what's done, state file locations, reconnect steps, remaining T-1.4/1.5,
Phase-2 platform prereq), infra/cluster.sh (status/start/stop/kubeconfig), and
document that tofu+ansible run on the host (libvirt provider limitation).
Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
Per operator rule: OpenTofu/age/sops/flux are Go binaries, so bake the toolchain
into a pinned Fedora-44 dev container (matches host libvirt 12) instead of a venv.
infra/devcontainer/: Containerfile + versions.env (tofu 1.9.0, age 1.2.1, sops
3.9.4, flux 2.4.0, kubectl 1.32.0, helm 3.16.0) + dev.sh (runs with host libvirt
socket + repo + ssh mounted) + README. Built & validated: all tools resolve;
virsh reaches host libvirt from inside the container.
Also record resolved Phase 1 decisions in PHASE1-SCOPE.md (6GiB/4vCPU x3;
k3s-dev NAT 10.73.0.0/24; container toolchain). Gate green.
Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
jakman85andClaude Opus 5.5