Kamaji vs vCluster Platform

Kamaji runs Kubernetes control planes as pods inside a management cluster, giving platform teams a way to operate multiple tenant clusters. Each tenant control plane is independent and separated from others. It uses upstream Kubernetes components and kubeadm, with CNCF-conformant tenant clusters. Worker nodes may run on bare metal, virtual machines, or cloud instances across cloud, data-center, edge, and bare-metal environments. Kamaji integrates with Cluster API for infrastructure provisioning and lifecycle management, and supports cluster provisioning, multi-cluster management, policy enforcement, and upgrade management. Datastores include etcd and MySQL- or PostgreSQL-compatible databases through kine; NATS support is experimental. Konnectivity, enabled by default, creates secure tunnels to worker nodes. Prometheus-format metrics cover management and tenant planes. Metrics endpoints use unauthenticated plaintext by default, with documented HTTPS and bearer-token options. Kamaji is free under Apache 2.0; Clastix also sells enterprise support.

vCluster Platform is a management plane for deploying, configuring, and operating tenant cluster fleets through a web UI, CLI, and API. It manages tenants across EKS, GKE, AKS, bare metal, and private cloud, with isolated tenant clusters and shared-node, dedicated-node, private-node, Auto Nodes, and Standalone tenancy models. With vCluster Cloud, the platform is hosted while host and virtual clusters remain in the customer's infrastructure; self-hosted deployment puts the full platform there. The Free plan is free and permits unlimited virtual clusters within a limit of 64 CPUs or 32 GPUs, with a maximum of one HA vCluster. It also supports provisioning, multi-cluster management, policy enforcement, and upgrade management. Listed integrations include Terraform, Rancher, Argo CD, HashiCorp Vault, and Istio. Enterprise features include SSO, audit logging, vulnerability alerts, security review, FIPS-compliant images, enterprise support, and SLAs. One isolation caveat is that shared nodes share several underlying components across tenants, including the kernel, kubelet, node filesystem, and host network. The maker describes use by AI cloud providers and public and private cloud teams.