Self-hosted Renovate as a weekly CronJob (Sun 04:00 Europe/Budapest)
opening dependency-update PRs against admin/homelab-manifests on Gitea.
Pilot is deliberately narrow:
- Only the kubernetes + helm-values managers are enabled.
- Default-deny packageRule; only four images may update:
opengist, uptime-kuma, gokapi, cal.com.
- minor/patch -> PR with Gitea native auto-merge (platformAutomerge).
- major -> held for manual approval via Dependency Dashboard checkbox.
Image pinned to renovate/renovate:43.197.0 (the plain tag is the
minimal image; the -slim suffix was retired upstream after v37.440.x).
Stateless: no Service/Ingress/PVC. Read-only root FS with a 2Gi /tmp
emptyDir for git clones + cache. Secrets from existing renovate-secrets.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
DISABLE_SIGNUP=true blocks OIDC auto-register too (callback returns 'Signups are
currently disabled by the administrator'), so the first admin account can't be
created. Set DISABLE_SIGNUP=false just long enough for the admin to log in via
Authentik once (auto-registers + admin via SPARKY_FITNESS_ADMIN_EMAIL); will be
reverted to true right after.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The Authentik application was created with slug 'sparky-fitness' (hyphen), so the
OIDC discovery document lives at /application/o/sparky-fitness/.well-known/...
The previous value (no hyphen) 404'd. Align the issuer URL with the actual slug.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
SparkyFitness (https://github.com/CodeWithCJ/SparkyFitness) replaces wger on
workout.dooplex.hu / workout.home with native Authentik OIDC.
Components (sparkyfitness.yaml): dedicated postgres:15-alpine, server (3010,
/api/health), frontend nginx (root image, listens :80). PVCs sparkyfitness-postgres
+ sparkyfitness-uploads (Longhorn, backup labels). In-app OIDC, no forward-auth.
Deviations from the deploy spec, following upstream ground truth:
- In-container mount path for uploads is /app/SparkyFitnessServer/uploads (per the
upstream Helm chart values.yaml), not /app/uploads.
- Frontend root image (codewithcj/sparkyfitness) has 'listen 80;' hardcoded ->
NGINX_LISTEN_PORT=80, containerPort/Service 80.
- Image names use the docker-compose variants (codewithcj/sparkyfitness_server,
codewithcj/sparkyfitness); the upstream Helm chart uses -server/-frontend.
- All wger Deployments scaled to 0 (incl. celery worker/beat), not just
wger + wger-redis, so celery doesn't crashloop against the downed redis.
Secrets (outside git): sparky-oauth (client-id/client-secret, pre-existing) +
sparky-app (db + app-db creds, api-encryption-key, better-auth-secret).
wger is parked, not deleted: Deployments at 0, both Ingresses removed (ArgoCD
prune frees the hostnames); Services/ConfigMap/PVCs/CNPG DB kept for rollback.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Runs node apps/api/dist/migrate.js after wait-for-db and wait-for-redis,
applying Drizzle SQL migrations before the API starts. Fixes fresh
deployment failing with "relation does not exist".
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
admin
Renovate Bot