You have a working Publisher. Where to from here?

Productionise the deployment

  • Switch to networking.mode: pod if you haven’t already. Lower
    blast radius, works on managed Kubernetes, no privileged host network
    required.
    → pod-vs-host network
  • Run an HA pair (two replicas across nodes, anti-affinity).
    → HA pair
  • Enable persistence so settings survive node failures.
    → State management
  • Pin the image tag to a tested version instead of latest.
    → BYO image

Roll out to your real cluster

  • EKS / AKS / GKE / OpenShift — see
    distributions for the
    per-platform tweaks (PSA, GKE node images, OpenShift SCC, etc.).
  • Multiple clusters / regions — one Helm release per cluster, one
    enrollment.commonName per cluster.
    → Multi-cluster

Operate it