fix(renovate): edit catalog sources, track CNPG imageName, honour update.range, guard real-run auto-merges

Audit of the first completed dry runs (353/354) plus a debug-level local reproduction: - App-image updates edited only the generated apps/wave-*. A ServiceDefinition has no apiVersion, so the kubernetes manager extracts nothing from catalog/, and every such PR would fail render-golden (reproduced: GOLDEN DRIFT). New regex managers for catalog source.image and render.py OAUTH2_PROXY_IMAGE put the source and generated edits in one branch (render-golden rc 0, reproduced). - CNPG Cluster imageName was untracked (3 Clusters), while the grant Job was tracked. New manager. - allowedVersions mirror catalog update.range: this drops gitea 28.x, qbittorrent 20.x, radarr 6, nextcloud 35 and postgresql 18. - Never auto-merge: gitea, act_runner, postgresql (blast radius), flux range widening, python minors. Majors wait for a Dependency Dashboard tick. - velero plugin RC: semver versioning (v1.14.3-rc.2 -> v1.14.4 now proposed). - flux docker duplicates disabled (the "Cannot find replaceString" line). Two false comments corrected. Local dry run: regex deps 7 -> 20, branches 33 -> 29. The validator passes on 44.167.0. Co-Authored-By: Claude Opus 5.5 <[email protected]>
2026-10-30 02:35:33 -05:00
1 parent 593d0fad15
commit d9571fc6d8
@@ -149,6 +149,50 @@
      ],
      "datasourceTemplate": "docker",
      "versioningTemplate": "docker"
+    },
+    // ADDED 2026-10-30 — THE CATALOG IS THE SOURCE; apps/wave-*/*.yaml IS GENERATED FROM IT.
+    // First real dry run (dispatch runs 353/354): EVERY app-image update edited only the rendered
+    // apps/wave-*/*.yaml and NONE edited catalog/. The `kubernetes` manager lists catalog/ above
+    // but extracts nothing there: a ServiceDefinition has no `apiVersion`, so it is not a
+    // Kubernetes manifest to that manager (measured: 0 deps from catalog/). Every such PR would
+    // fail ci/render-golden.sh (the canonical render reverts the bump), sit red and hold a
+    // prConcurrentLimit slot. These two managers read the SOURCES, under the SAME depName as the
+    // rendered line, so both land in ONE branch and render==golden holds: `source.image` in each
+    // ServiceDefinition, and render.py's OAUTH2_PROXY_IMAGE (the six auth sidecars).
+    {
+      "customType": "regex",
+      "description": "ServiceDefinition `source.image: <repo>:<tag>` (catalog/*.yaml) — the source of the rendered apps/wave-* image.",
+      "managerFilePatterns": ["/(^|/)catalog/[^/]+\\.ya?ml$/"],
+      "matchStrings": [
+        "\\n\\s+image:\\s*(?<depName>[^\\s:@#\"']+):(?<currentValue>[^\\s@#\"']+)"
+      ],
+      "datasourceTemplate": "docker"
+    },
+    {
+      "customType": "regex",
+      "description": "catalog/render.py OAUTH2_PROXY_IMAGE — the per-app oauth2-proxy sidecar every forward-auth app is rendered with.",
+      "managerFilePatterns": ["/(^|/)catalog/render\\.py$/"],
+      "matchStrings": [
+        "\\nOAUTH2_PROXY_IMAGE\\s*=\\s*\"(?<depName>[^\":@]+):(?<currentValue>[^\"@]+)\""
+      ],
+      "datasourceTemplate": "docker"
+    },
+    // ADDED 2026-10-30 — CNPG `Cluster.spec.imageName`. The `kubernetes` manager reads `image:` only,
+    // so the three database Clusters (platform/keycloak/postgres-cluster.yaml, apps/wave-0/
+    // nextcloud-db.yaml, apps/wave-0/semaphore-db.yaml) were untracked, while the grant Job in
+    // nextcloud-db.yaml — whose comment says "keep the tag in step with spec.imageName" — WAS. A real
+    // run would have bumped the Job and left the database behind. Same depName, so one branch.
+    {
+      "customType": "regex",
+      "description": "CloudNativePG Cluster `imageName: <repo>:<tag>`.",
+      "managerFilePatterns": [
+        "/(^|/)platform/.+\\.ya?ml$/",
+        "/(^|/)apps/.+\\.ya?ml$/"
+      ],
+      "matchStrings": [
+        "\\n\\s+imageName:\\s*(?<depName>[^\\s:@#\"']+):(?<currentValue>[^\\s@#\"']+)"
+      ],
+      "datasourceTemplate": "docker"
    }
  ],
  "packageRules": [
@@ -200,7 +244,15 @@
        "longhorn",
        "cilium",
        "cloudnative-pg",
-        "kube-prometheus-stack"
+        "kube-prometheus-stack",
+        // ADDED 2026-10-30. gitea and act_runner ARE the merge gate: a bad auto-merged bump takes
+        // down the very check that would have to approve its revert (the dry run proposed gitea
+        // 1.24.3 -> 1.27.3 and act_runner 0.2.11 -> 0.6.1, both as auto-mergeable "minor"; a 0.x
+        // minor is a breaking line by semver). The postgresql image restarts every CNPG database,
+        // Keycloak's (the IdP) included, now that imageName is tracked.
+        "gitea/gitea",
+        "gitea/act_runner",
+        "ghcr.io/cloudnative-pg/postgresql"
      ],
      "automerge": false,
      "addLabels": [
@@ -227,6 +279,51 @@
        "gitea.home.keithjackson.dev/home-network/conformance-runner"
      ],
      "addLabels": ["own-image"]
+    },
+    {
+      "description": "ADDED 2026-10-30. The `flux` manager also extracts `image:` lines inside HelmRelease values, which the `kubernetes` manager already owns. Measured: exactly two (alpine in platform/oauth2-proxy/helmrelease.yaml, velero-plugin-for-aws in platform/velero/helmrelease.yaml), each extracted twice; the second rewrite of the same line logged `Cannot find replaceString in current file content` in runs 353/354. Flux keeps its charts (datasource helm).",
+      "matchManagers": ["flux"],
+      "matchDatasources": ["docker"],
+      "enabled": false
+    },
+    {
+      "description": "ADDED 2026-10-30. MIRRORS catalog/<svc>.yaml `update.range` — the operator's declared per-service policy (schemas/servicedefinition.schema.json), which Renovate does not read. Without it the dry run proposed gitea 28.2.0, qbittorrent 20.04.1 (both stale/odd tags, not real successors), radarr 6.x and nextcloud 35 (four majors past 31; Nextcloud cannot skip majors). KEEP IN STEP with the catalog ranges.",
+      "matchPackageNames": ["gitea/gitea"],
+      "allowedVersions": "<2.0.0"
+    },
+    { "matchPackageNames": ["nextcloud"], "allowedVersions": "<32.0.0" },
+    { "matchPackageNames": ["lscr.io/linuxserver/qbittorrent"], "allowedVersions": "<6.0.0" },
+    { "matchPackageNames": ["lscr.io/linuxserver/radarr"], "allowedVersions": "<6.0.0" },
+    { "matchPackageNames": ["lscr.io/linuxserver/sonarr"], "allowedVersions": "<5.0.0" },
+    { "matchPackageNames": ["semaphoreui/semaphore"], "allowedVersions": "<3.0.0" },
+    {
+      "description": "ADDED 2026-10-30. A CNPG Postgres MAJOR is a migration (new cluster + import), not an image bump; the dry run proposed 16.4 -> 18.6. Minors stay (and are blast-radius, above).",
+      "matchPackageNames": ["ghcr.io/cloudnative-pg/postgresql"],
+      "allowedVersions": "<17.0.0"
+    },
+    {
+      "description": "ADDED 2026-10-30. A HelmRelease semver RANGE is a policy ceiling, and Renovate does not respect it — it WIDENS it (dry run: external-dns `<1.22.0` -> `<1.24.0`, headlamp `<0.42.0` -> `<0.47.0`), and the default rule would then auto-merge that as a 'minor', silently letting Flux adopt every new minor. Raising a ceiling is deliberate per-component work: propose it, never auto-merge it.",
+      "matchManagers": ["flux"],
+      "matchCurrentValue": "/^[<>=~^]|\\.x$/",
+      "automerge": false,
+      "addLabels": ["needs-review", "range-widen"]
+    },
+    {
+      "description": "ADDED 2026-10-30. velero-plugin-for-aws is pinned to a RELEASE CANDIDATE (v1.14.3-rc.2) and Renovate proposed NOTHING for it, although v1.14.3 and v1.14.4 are stable on Docker Hub. Docker versioning treats everything after the first `-` as a compatibility SUFFIX, so only other `*-rc.2` tags qualify and the RC is a fossil forever. Semver versioning reads it as a prerelease, and a prerelease upgrades to the stable release.",
+      "matchPackageNames": ["velero/velero-plugin-for-aws"],
+      "versioning": "semver"
+    },
+    {
+      "description": "ADDED 2026-10-30. A python MINOR is a language-runtime release (the dry run proposed 3.13 -> 3.15 for the Keycloak reconcilers), and the gate does not execute those scripts. Review it; patches still auto-merge.",
+      "matchPackageNames": ["python"],
+      "matchUpdateTypes": ["minor"],
+      "automerge": false,
+      "addLabels": ["needs-review"]
+    },
+    {
+      "description": "ADDED 2026-10-30. MAJORS WAIT FOR A TICK ON THE DEPENDENCY DASHBOARD instead of opening a PR. They are never auto-merged anyway, and with prConcurrentLimit 5 the dry run's 10 majors would hold every PR slot open indefinitely, starving the patch/minor/digest PRs that can merge on green.",
+      "matchUpdateTypes": ["major"],
+      "dependencyDashboardApproval": true
    }
  ],
  // WHAT RENOVATE STILL CANNOT FIX, so nobody mistakes a working bot for full coverage. A
@@ -235,10 +332,12 @@
  // 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.
+  // nothing recording what is running. CORRECTED 2026-10-30: nextcloud:31-apache and docker:27-dind
+  // ARE maintained (the first dry run proposed 35-apache and 29-dind; the suffix is kept).
+  // Separately, the HelmRelease minor-locking ranges (9 of 12, e.g. `>=88.0.0 <89.0.0` on
+  // kube-prometheus-stack): THIS FILE USED TO SAY they cap Renovate at patches. They do not —
+  // Renovate WIDENS the range (runs 353/354). The `range-widen` rule above keeps that a reviewed
+  // PR rather than an auto-merge. Both lists are in decisions/2026-10-17-automated-upgrades.md.
  // 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