Everything you can build, run and use on GRN.CLOUD

Vanilla Kubernetes (OKE), open-source OpenShift (OKD) or Red Hat OpenShift (OCP). CNCF-conformant in every case.

HyperShift runs the API server, etcd and console as pods. No control-plane nodes to operate — or to pay for.

KubeVirt VM workers on VPC, or bare-metal workers on DPC. Pools are sized and scaled independently.

Multi-tenancy with roles, role bindings, service accounts and per-namespace resource limits.

Gateway API for hostname and path routing; MetalLB provides LoadBalancer services on bare metal, advertised with BGP.

Topology spread across three availability zones, pod disruption budgets and readiness-gated rolling updates.

CSI-provisioned Ceph and OpenEBS volumes with snapshots and online expansion, requested by PersistentVolumeClaim.

Provisioning, version upgrades, security patching and backup of the control plane and nodes.

Raw volumes with exclusive access and the shortest path to disk. Databases, message brokers and VM disks.

A shared POSIX filesystem mounted read-write by many pods across nodes at once. Shared document roots and upload directories.

S3-compatible buckets over HTTPS, reachable in or out of the cluster. Backups, artifacts, datasets and media.

Low-latency local volumes for the most I/O-sensitive workloads, where latency matters more than shared placement.

Ceph replicates volumes across availability zones, so a zone can be lost without losing data.

Dynamic provisioning, snapshots and clones, and online volume expansion without downtime.

Kubernetes NetworkPolicy enforced in the data plane. A default-deny namespace admits only the connections it declares.

MetalLB assigns external addresses to LoadBalancer services and advertises them with BGP; routers spread flows using ECMP.

The same service address is announced from more than one location; withdrawn announcements converge traffic onto healthy paths.

Gateway API routing, cluster DNS for service discovery, cert-manager for certificates and External-DNS for records.

VXLAN private subnets, and user-defined networks (UDN) that give virtual machines their own layer-2 or layer-3 segments.

OpenVPN and WireGuard run as ordinary cluster workloads, for remote administrative access and site-to-site links.

Submariner and BGP EVPN connect clusters across zones, sites and on-premises, with cross-cluster service discovery.

Three independent zones per regional site, each with its own power, cooling, uplinks and network fabric.

Jupyter notebooks from curated, versioned images, isolated per user, with data connections configured up front.

Kubeflow Pipelines turn interactive notebook code into repeatable, distributed training runs.

KServe inference endpoints, ModelMesh for density, and canary rollout with rollback.

Versioned models tracked from experiment through to the endpoint serving them.

NVIDIA GPUs, MIG-partitionable, requested like any other resource so capacity is shared rather than stranded.

Datasets and checkpoints on S3-compatible storage, and inference served from infrastructure in the Netherlands.

The primary interface. Every workload, volume, route and policy is a resource you can create, inspect and version.

Cluster, workload, virtual machine and storage management in the browser, for the things that are faster to click.

Standard kubectl and oc against your cluster, with no wrapper tool in between.

Chart-based packaging and release management for your own and third-party applications.

Argo CD is provided and operated; Flux is supported as a customer-installed alternative.

Model Context Protocol access, so AI assistants and agents can work against the platform directly.