k8s-and-chill/rendered/envs/minikube/ocis/cronjob-thumbnails-cleanup.yaml
Felix Wolf fe51c8c1bc feat(minikube): add minikube environment with garage S3 backend
Adds a self-contained minikube environment for local development and
testing alongside the existing production env.

env: minikube
  - cluster.domain: minikube (browser DNS routes *.minikube → minikube ip)
  - tls issuer: mkcert (CA-signed via cert-manager mkcert ClusterIssuer)
  - storageClass: standard (minikube hostpath provisioner)
  - backups disabled; storagebox disabled
  - excludes argocd, forgejo, hcloud-csi (manual kubectl apply for testing)

prototypes/garage:
  - hand-rolled S3-compatible object store (single Deployment + PVC)
  - mittwald-generated rpc_secret + admin_token (hex)
  - PostSync init Job: assigns cluster layout, ensures bucket and access
    key, writes ocis-s3-credentials cross-namespace into ocis ns
  - idempotent: skips if k8s secret already populated; otherwise rotates
    the garage key (admin API only returns secretAccessKey on create)
  - cross-ns RBAC re-pinned via zz-cross-ns-rbac-fix overlay (ns.ytt.yaml
    clobbers explicit namespace fields)

ocis:
  - new admin-user-id init Job ensures secret.user-id is a valid UUID v4
    (mittwald can't generate UUIDs; ocis-settings rejects non-UUID ids)
  - mittwald no longer manages user-id; existing prod UUIDs preserved
  - insecure flag (oidcIdpInsecure / ocisHttpApiInsecure / ocmInsecure)
    parameterized; defaults to false; minikube sets true for self-signed
    OIDC issuer URL trust

other prototypes:
  - victoria-metrics-single helm values ytt-ified (storageClassName)
  - grafana admin secret now generated by mittwald (was hand-created in
    prod; manifest is no-op there since mittwald only fills empty fields)

flake.nix: minikube + docker + postgresql added to dev shell.
2026-05-03 17:23:57 +02:00

73 lines
2.3 KiB
YAML

apiVersion: batch/v1
kind: CronJob
metadata:
annotations:
a8r.io/repository: ssh://git@git.tr1ceracop.de:222/gitea_admin/k8s-and-chill.git
ignore-check.kube-linter.io/latest-tag: using the stable tag on this busybox image is better than having an outdated image
labels:
app.kubernetes.io/instance: ocis
app.kubernetes.io/managed-by: Helm
app.kubernetes.io/name: ocis
app.kubernetes.io/version: 7.1.4
helm.sh/chart: ocis-0.7.0
name: thumbnails-cleanup
namespace: ocis
spec:
concurrencyPolicy: Forbid
failedJobsHistoryLimit: 3
jobTemplate:
spec:
parallelism: 1
template:
metadata:
labels:
app: thumbnails-cleanup
app.kubernetes.io/instance: ocis
app.kubernetes.io/managed-by: Helm
app.kubernetes.io/name: ocis
app.kubernetes.io/version: 7.1.4
helm.sh/chart: ocis-0.7.0
spec:
containers:
- command:
- /bin/sh
- /script/thumbnails-cleanup-script.sh
image: busybox:stable
imagePullPolicy: IfNotPresent
name: thumbnails-cleanup
resources: {}
securityContext:
allowPrivilegeEscalation: false
capabilities:
drop:
- ALL
readOnlyRootFilesystem: true
runAsGroup: 1000
runAsNonRoot: true
runAsUser: 1000
volumeMounts:
- mountPath: /var/lib/ocis
name: thumbnails-data
- mountPath: /script
name: script
nodeSelector: {}
restartPolicy: Never
securityContext:
fsGroup: 1000
fsGroupChangePolicy: OnRootMismatch
seccompProfile:
type: RuntimeDefault
volumes:
- name: thumbnails-data
persistentVolumeClaim:
claimName: thumbnails-data
- configMap:
defaultMode: 320
name: thumbnails-cleanup-script-configmap
name: script
ttlSecondsAfterFinished: 86400
schedule: 0 * * * *
startingDeadlineSeconds: 600
successfulJobsHistoryLimit: 3
suspend: true