Files
home-network-k3s-impl/renovate.json
T
jakman85andClaude Opus 5 a2440a3125
gate / gate (pull_request) Successful in 1m57s
gate / conformance (pull_request) Has been skipped
feat(upgrades): automated OS reboots (kured) + security-only unattended-upgrades, and give Renovate a runner
Two halves of one repeatable upgrade process. Neither works alone.

(a) OS upgrades and reboots

platform/kured/ — Kubernetes Reboot Daemon, chart 6.1.0/app 1.23.0, in its own
Flux Kustomization (dependsOn platform, wait:true). NOT in platform/kustomization.yaml:
that tier gates platform-config and every apps wave, so a wedged reboot daemon there
could stall every application. A component added to make maintenance safer must not be
able to cause an outage.

infra/ansible/roles/unattended_upgrades/ — new role (node_baseline is touched by other
open branches, and this is a different lifecycle), enabled for prod in group_vars.

The human gate is the SENTINEL, which is how this honours node_baseline's "IT DOES NOT
REBOOT" (k3s-prod-2 hung on a warm reboot 2026-09-03) and cluster-upgrade-plan.md's
rejection of automating away the gate between nodes: u-u is security-only with the
kernel blacklisted, so it essentially never writes /var/run/reboot-required. That file
appears when an operator deliberately runs node-baseline.yml --tags kernel. kured then
does only the mechanical half, including the ordering rule node_baseline has to shout.

Quorum: concurrent reboots are structurally impossible, not merely unlikely.
concurrency:1 plus kured's single lock annotation (API-server optimistic concurrency),
and lockTtl:0 — an expiring lock is exactly the 3-node etcd catastrophe, so a hung node
holds it forever and nothing cascades.

Longhorn: measured 25 volumes x 3 replicas, hard anti-affinity, 600s replenishment wait,
serialized rebuild — so every drain degrades ALL volumes to 2/3 and the inter-node gate
must be volume robustness, not pod readiness. Doubled: lockReleaseDelay 30m
(unconditional) plus kured's Prometheus alert gate (semantic).

alertFilterMatchOnly:true is load-bearing — the flag inverts the regexp's meaning, and
under the chart default kured would have been blocked on day one by the five alerts
measured firing with no end in sight, and never rebooted anything.

Kernel floor (kernel >= 6.13, EFI memattr): excluded TWICE. The running kernel comes
from o=Debian Backports, which is not an allowed origin, so u-u cannot see a kernel
upgrade; and ^linux- is blacklisted anyway so widening origins later cannot re-arm it.
Asserted against apt-config dump, not against the file just written. 6.12 kernels are
still installed and bootable on all three nodes, and nothing was held.

Closes a silent no-op: update-notifier-common was absent on all three nodes, so NOTHING
could create the sentinel. kured alone would have idled forever at 1/1 Running.

platform/kured/alerts.yaml — three rules over kured's only metric, with a companion
promtool suite pinning each both ways, plus runbooks/node-reboot-automation.md.
Window is DAILY 03:00-06:00 because a weekly one forces for:8d, longer than this
cluster's 7d Prometheus retention.

(b) Image/dependency automation

Renovate was aspirational: renovate.json present and complete, with no workflow, no
in-cluster workload, and its docker+kubernetes managers disabled — so every plain
image: tag was maintained by nothing while seven files claimed otherwise.

.gitea/workflows/renovate.yaml — scheduled daily, dry_run default true for a safe
first run, and a preflight that FAILS (not skips) without RENOVATE_TOKEN, because a
skipped Gitea job reports success and that is how this stayed inert.

Renovate now owns both halves; Flux image-automation is formally closed: it pushes
commits rather than opening PRs (bypassing CLAUDE.md §1), and it cannot work here
anyway — main is protected user_can_push:false and the flux-system GitRepository has
no secretRef. Its controllers are installed and have idled 43 days with zero CRs.

conformance/: +34/-0, one new SCRAPE_OWNERS entry claiming kured_. Weakens nothing;
tightens the check for this component (requires service.create+metrics.create, both
chart-default false, plus a companion NetworkPolicy requirement). Required because
assertion (3) fails an alert over an unclaimed metric and KNOWN_UNWIRED_METRICS is
shrink-only.

Sensitive paths touched: clusters/**, conformance/**, .gitea/** — needs a non-author
approving review; /approve-override deliberately NOT posted.

NOT verified: no OS upgrade run, no node rebooted, no lock observed, the Ansible role
never executed, and whether CNPG switches its primary on a drain-induced eviction (the
most likely way the first real reboot fails). Full list in the handoff entry.

Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
2026-10-18 00:10:29 -05:00

100 lines
8.5 KiB
JSON

{
"$schema": "https://docs.renovatebot.com/renovate-schema.json",
"_comment": "Renovate config — Phase 4 automated updates (ADR-0009, R-UPD-01/02). AS OF 2026-10-17 RENOVATE OWNS BOTH HALVES: Helm chart/dependency bumps AND per-service container image tags. It opens PRs for everything; allowed low-risk classes auto-merge on green CI. The ADR-0009 split that gave image tags to Flux image-automation is formally closed — see .gitea/workflows/renovate.yaml for the three measured reasons (it pushes commits rather than opening PRs, so it bypasses the CLAUDE.md §1 gate; `main` is protected with user_can_push:false and the flux-system GitRepository has no secretRef, so it cannot write anyway; and enabling it would need an in-cluster git WRITE credential for the repo that defines the cluster). image-automation has been commented out at platform/kustomization.yaml:52 since it was authored and the cluster holds zero ImagePolicy/ImageRepository/ImageUpdateAutomation objects, so nothing is being torn down. CONSEQUENCE: the `docker` and `kubernetes` managers, previously disabled ONLY to avoid fighting that controller, are now ENABLED — without them every plain `image: repo/name:tag` line in this repo was maintained by nothing at all, which is how the oauth2-proxy auth sidecars reached 7 minors behind the oauth2-proxy running in the same cluster. RUNNER: .gitea/workflows/renovate.yaml (scheduled daily); platform/endpoint are set there, not here.",
"extends": [
"config:recommended",
":dependencyDashboard",
":semanticCommits",
"helpers:pinGitHubActionDigests"
],
"timezone": "America/New_York",
"schedule": ["after 3am and before 6am"],
"prConcurrentLimit": 5,
"prHourlyLimit": 2,
"commitMessagePrefix": "chore(deps):",
"labels": ["dependencies", "automated-update"],
"ignoreUnstable": true,
"respectLatest": true,
"enabledManagers": ["flux", "kustomize", "helmv3", "kubernetes", "regex", "github-actions"],
"flux": {
"managerFilePatterns": ["/(^|/)platform/.+\\.ya?ml$/", "/(^|/)apps/.+\\.ya?ml$/"]
},
"kustomize": {
"managerFilePatterns": ["/(^|/)kustomization\\.ya?ml$/"]
},
"_kubernetes_comment": "ADDED 2026-10-17. The `kubernetes` manager reads plain `image:` fields out of Kubernetes manifests, and it is NOT auto-detected — it only looks at files matched here, which is why this block is as broad as it is. Scope covers every place this repo writes a container image: platform components, the rendered app waves, the ServiceDefinition catalog that generates them, the runbooks web server and the pm-agent job template. `catalog/` matters as much as `apps/`: apps/wave-*/*.yaml are GENERATED by catalog/render.py, so a bump applied only to the rendered output is reverted by the next render and would also fail ci/render-golden.sh. Renovate will raise BOTH, and both must land together.",
"kubernetes": {
"managerFilePatterns": [
"/(^|/)platform/.+\\.ya?ml$/",
"/(^|/)platform-config/.+\\.ya?ml$/",
"/(^|/)apps/.+\\.ya?ml$/",
"/(^|/)catalog/.+\\.ya?ml$/",
"/(^|/)runbooks/serving/.+\\.ya?ml$/",
"/(^|/)infra/agent/.+\\.ya?ml$/"
]
},
"customManagers": [
{
"customType": "regex",
"description": "Inline `# renovate: datasource=... depName=...` markers on pinned URLs / version literals (e.g. platform/kubevirt/kustomization.yaml release URLs, ci/versions.env, infra/devcontainer/versions.env).",
"managerFilePatterns": [
"/(^|/)kustomization\\.ya?ml$/",
"/(^|/)versions\\.env$/"
],
"matchStrings": [
"#\\s*renovate:\\s*datasource=(?<datasource>\\S+)\\s+depName=(?<depName>\\S+)(?:\\s+versioning=(?<versioning>\\S+))?\\s*\\n.*?(?<currentValue>v?[0-9]+\\.[0-9]+\\.[0-9]+)"
],
"versioningTemplate": "{{#if versioning}}{{{versioning}}}{{else}}semver{{/if}}"
}
],
"packageRules": [
{
"description": "Auto-merge the allowed low-risk classes on green CI — no human authoring (R-UPD-02). The Gitea merge gate (make -C ci gate) is the required check.",
"matchUpdateTypes": ["minor", "patch", "digest"],
"automerge": true,
"automergeType": "pr",
"platformAutomerge": true
},
{
"description": "MAJOR / potentially-breaking bumps require human review (docs/11 safety rail). Never auto-merged.",
"matchUpdateTypes": ["major"],
"automerge": false,
"addLabels": ["needs-review", "breaking"]
},
{
"description": "Keep the KubeVirt + CDI operator pair on a compatible track — grouped so they bump together (platform/kubevirt/kustomization.yaml).",
"matchDepNames": ["kubevirt/kubevirt", "kubevirt/containerized-data-importer"],
"groupName": "kubevirt-cdi"
},
{
"description": "Pinned tool versions (ci/versions.env, infra/devcontainer/versions.env) are a deliberate change surface — label but do not auto-merge; a version bump there changes CI reproducibility.",
"matchFileNames": ["ci/versions.env", "infra/devcontainer/versions.env"],
"automerge": false,
"addLabels": ["ci-tooling"]
},
{
"description": "ADDED 2026-10-17. NEVER AUTO-MERGE THE BLAST-RADIUS COMPONENTS, not even a patch. The default rule above auto-merges minor/patch/digest on a green gate, which is right for an app image: a bad Sonarr bump inconveniences one pod. It is wrong for these five, because the gate is a STATIC check — it renders manifests and lints them, it does not boot a node, drain a volume or form an etcd quorum (vars.CLUSTER_AVAILABLE is unset, so L2/L3 conformance does not run in CI at all). So for anything whose failure mode is the CLUSTER rather than a service, a green gate is not evidence and a human must read the changelog. kured is here because it reboots control-plane nodes; longhorn because a storage-engine bump migrates live volume data; cilium because it is the CNI and a bad bump is a total partition (and its version is DUAL-PINNED in infra/ansible/roles/cilium/defaults/main.yml, which Renovate does not see, so the two would silently desync); cloudnative-pg because it drives Postgres switchovers; kube-prometheus-stack because ci/versions.env asserts an exact promtool/Prometheus lockstep and a chart bump that moves Prometheus fails the gate until that file moves with it.",
"matchPackageNames": [
"kured",
"longhorn",
"cilium",
"cloudnative-pg",
"kube-prometheus-stack"
],
"automerge": false,
"addLabels": ["needs-review", "blast-radius"]
}
],
"_unmanageable_comment": "WHAT RENOVATE STILL CANNOT FIX, so nobody mistakes a working bot for full coverage. A FLOATING tag has no version to compare, so Renovate skips it silently — it does not pin it and does not warn. As of 2026-10-17 that is: plexinc/pms-docker:latest, lscr.io/linuxserver/jackett:latest, pihole/pihole (NO TAG AT ALL), quay.io/keycloak/keycloak (no tag), nextcloud:31-apache and docker:27-dind. Those must be pinned BY HAND to a concrete version before any automation can maintain them, and until then they drift on every pull with nothing recording what is running. Separately, the HelmRelease minor-locking ranges (9 of 12, e.g. `>=88.0.0 <89.0.0` on kube-prometheus-stack) cap Renovate at patches inside a line it has been told not to leave; raising those ceilings is deliberate per-component work and is NOT part of enabling the bot. Both lists are in decisions/2026-10-17-automated-upgrades.md.",
"ignorePaths": [
"**/*.example",
"**/*.sops.yaml",
"conformance/**",
"platform/image-automation/per-service/**",
"platform/kyverno/tests/**"
],
"_ignorePaths_comment": "platform/kyverno/tests/** ADDED 2026-10-17: those Kyverno policy fixtures pin `alpine:latest` and a `conformance-runner:latest`, which are test inputs whose whole point is to be a fixed shape, not software to keep current. `conformance/**` was already ignored for the same reason; this closes the matching hole that enabling the `kubernetes` manager would otherwise have opened.",
"_docker_comment": "The `docker` manager (Dockerfiles, docker-compose) and the `kubernetes` manager above were BOTH previously `enabled: false`, solely so Renovate would not fight Flux image-automation over a running tag. That controller was never enabled and never will be here (see _comment), so the only thing those two flags were achieving was leaving every plain `image:` tag in the repo unmanaged. `docker` is left out of enabledManagers because this repo contains no Dockerfile that ships to the cluster; `kubernetes` is now ON and is the one that matters.",
"docker": { "enabled": true }
}