Kubernetes features and limits

_{Updated August 24, 2026}

Limits

Limit	Value
Clusters per account	10 (contact support to raise)
Node pools per cluster	unlimited; ≥1 required
Nodes per pool	1–20 (≥2 workers per cluster across all pools)
Minimum workers per cluster	2
Persistent volumes	`raff-block` default StorageClass — PVCs become Raff Volumes (10 GB–10 TB each, ReadWriteOnce)
MetalLB address pool	up to 32 addresses
Kubernetes versions	up to three minor versions with in-place upgrades — see Supported versions
Region	us-east
Connections / requests / Ingress hosts	No platform-imposed limits — throughput is bounded only by your cluster’s dedicated gateway capacity (per-cluster, unaffected by other tenants)

Managed, free control plane (optional 3-master HA), RKE2 + Cilium
Node pools with labels, taints and autoscaling (all pools share the cluster’s worker plan)
Drain-first scale-down (PodDisruptionBudgets honoured, never force-deleted)
Admin kubeconfig plus short-lived kubeconfigs (10 minutes to 30 days) with one-click revocation; full REST API; dashboard with workloads, pods, live logs, metrics and control-plane health
App Hub one-click installs with HTTPS/TCP expose
Dedicated VPC, gateway firewall (baseline or your security group), Traefik and MetalLB options, per-GB volume-backed persistent storage
99.90% monthly SLA (endpoint path on HA clusters) and 24/7 support

Honest list of gaps we’re working on: