Product

Managed Kubernetes

Create and operate clusters on fabric you control — CapA Machines on Talos guests, Kamaji dense control planes, Cilium as the product CNI/LB path. Prime management, tenants, and imported kubeconfigs share one Cloud inventory.

Not a DIY Cluster API maze — ACM owns CapA document generation so desired state stays coherent.

Architecture

Prime, Tenant, and Imported on one fabric

Day-0 lean → move → scale-ha. Tenants get Kamaji + CapA workers. Tenant Kubernetes rolls use the per-cluster wizard; Prime stays on the fabric path — not that wizard.

Prime

Site management cluster. CapA control-plane and worker MachineDeployments. Kubernetes upgrades are a platform operation — not the tenant wizard.

Tenant

Kamaji TenantControlPlane plus CapA worker pools on Talos guests. Cilium VIP for the API. Scale pools and run the Kubernetes upgrade wizard here.

Imported

1
Day-0 shape
StartPlatform brings lean bootstrap → move management → scale-ha to the desired CP/worker replicas. Cilium and Kamaji land from the product path.
2
CapA + ACM
Cluster API Provider Aanjaleya places Machines on fabric hosts. ACM owns CapA document generation (Force-SSA) so desired state stays coherent.
3
Operate from Cloud
RegisteredCluster inventory shows kind, phase, version, and topology glyphs. Create tenant, import, scale pools, download kubeconfig, upgrade tenants.
4
Later rolls
Tenant Kubernetes upgrades use the per-cluster wizard (control plane → rolling workers → healthy). Prime stays outside that wizard — fabric day-0 / site ops.

Cluster ARNs

Deterministic product ARNs: aanjaleya:cluster:::<kind>:<name>. Minted via libraries — never hand-rolled string concat.

RegisteredCluster

apiVersion: cloud.aanjaleya.io/v1alpha1
kind: RegisteredCluster
metadata:
  name: crest-cicd
  namespace: tenant-crest-cicd
spec:
  kind: tenant
  displayName: crest-cicd
  endpoint: https://192.168.130.240:6443
  kubernetesVersion: v1.32.2
  arn: aanjaleya:cluster:::tenant:crest-cicd
  tenantControlPlaneRef:
    name: crest-cicd
    namespace: kamaji-system

Always-on core

Not an optional pack

Managed Kubernetes is platform core — CapA, Kamaji, Cilium, Talos Image Factory, and Cloud Clusters — not an optional pack toggle like Managed Postgres or Valkey.

Day-0 owns CNI and providers in-process. Manual cilium / helm / kubectl fixes during ensure-platform are out of policy.

See also Talos guests and air-gapped platform.

Open source

Cluster API

Kubernetes cluster lifecycle API. CapA is the Aanjaleya provider that places Machines on fabric hosts.

Kubernetes® icon (Cluster API is a Kubernetes SIG project)

cluster-api.sigs.k8s.io

Kamaji

Dense tenant control planes as pods — not a full CP VM trio per cluster by default.

github.com/clastix/kamaji

Cilium

Product CNI and LoadBalancer path for API VIPs. Manual host cilium/helm installs during ensure-platform are out of policy.

cilium.io

Talos Linux

Immutable guest OS for every node. Upgrades are image pins from Image Factory — not SSH patch loops.

www.talos.dev

Aanjaleya Innovation Private Limited is not affiliated with or endorsed by those projects in this attribution.

Talk to us Talos guests Air-gapped Managed Registries Block storage All features