Skip to main content

Managed Kubernetes

Run containers on a real Kubernetes cluster without building and babysitting the control plane. You choose a cluster size and a node shape; The Wahda Cloud provisions the control plane, joins the worker nodes, and hands you a kubeconfig. From there it's standard Kubernetes — kubectl, Helm, your existing manifests, your existing CI.

The cluster runs on the same platform as your VMs, in the same project, on the same private networks. That means a Kubernetes workload can talk to a managed database or a VM over the private network with nothing in between, and can expose itself to the internet through the same managed load balancer you'd use for a fleet of VMs.

Nothing about the cluster is proprietary. It speaks the standard Kubernetes API, so anything that works against upstream Kubernetes — kubectl, Helm charts, operators, GitOps controllers — works here.


What you get

CapabilityDetails
Managed control planeWe run and patch the API server, scheduler, controller manager, and cluster datastore. You don't get a bill for control-plane VMs and you don't get paged when one reboots.
Standard Kubernetes APIFull kubectl access via a downloadable kubeconfig. No forked API, no vendor CLI required.
Node poolsGroups of identically-shaped worker nodes. Add pools for different workloads, scale each pool independently. See Node pools.
Up to 100 nodes per clusterThe product ceiling for a single cluster. Real-world sizing is bounded by your project allocation and regional capacity — see Capacity and honest limits.
Private networkingNodes sit on a private network and subnet inside your project, alongside your VMs and databases.
Load balancer integrationPublish a Service or an ingress through the managed L4/L7 load balancer and a floating IP. See Cluster networking.
Persistent volumesPods get durable storage from the same SSD block-storage pool that backs VM disks.
India data residencyControl plane and nodes run in the North India region (in-north-az1), in our Faridabad data centre.
Console-drivenCluster lifecycle is managed from console.thewahda.com, like every other resource.
INR + GST billingPay-as-you-go, metered hourly, invoiced in INR with GST.

How the pieces fit

  • Cluster — the tenant boundary for Kubernetes. One managed control plane, one or more node pools, one private network.
  • Control plane — managed by us. You never SSH into it and you aren't billed for it as compute.
  • Node pool — a set of worker nodes that all share one flavor, one image, and one availability zone. A cluster can have several.
  • Node — a worker VM that runs your pods. Nodes consume your project's compute allocation exactly like any other instance.
  • Persistent volume — a block-storage volume attached to a node and mounted into a pod.
  • Service / ingress — how traffic reaches your pods. Cluster-internal traffic stays on the private network; public traffic comes in through a load balancer and a floating IP.

When to use managed Kubernetes (and when not to)

Use managed Kubernetes when

  • You already ship containers and have manifests or Helm charts you want to run as-is.
  • You want rolling deploys, self-healing, and horizontal scaling without wiring them together yourself out of systemd units and a reverse proxy.
  • You run many small services and want them to share a pool of machines instead of each owning a VM.
  • You want to scale worker capacity up and down as load changes, without rebuilding the platform underneath.
  • Your team already knows kubectl and doesn't want to learn a provider-specific orchestration layer.

Run plain VMs instead when

  • You have one or two services and no container pipeline. A pair of VMs behind a load balancer is simpler, cheaper, and easier to debug at 3 a.m.
  • Your workload is a stateful single process — a database, a message broker, a legacy monolith with local disk state. Use a VM, or a managed database where the engine is one we run.
  • You need host-level control — custom kernels, perf, hugepages, specific NIC tuning.
  • Nobody on the team owns Kubernetes. An unowned cluster is a liability. Kubernetes moves the operational burden, it doesn't remove it.

The honest rule: Kubernetes pays for itself when you have enough services that scheduling them by hand has become the bottleneck. Below that line, VMs win.


Capacity and honest limits

A single cluster supports up to 100 nodes. That is the platform ceiling, not a promise that any given cluster can reach it on demand — two things bound it in practice:

  1. Your project allocation. Worker nodes draw on the same vCPU, RAM, and storage allocation as your VMs. A 100-node cluster of even modest nodes is a large allocation request. See Projects & quotas.
  2. Regional capacity. The North India region is a finite, physical cluster. Very large clusters — and RAM-heavy node shapes in particular — need to be planned with us rather than provisioned blind.

If you're sizing a cluster beyond a handful of nodes, or you need a shape that isn't in the published flavor catalog, email info@thewahda.com before you build it. We'll confirm capacity and reserve it. That conversation takes a day; discovering a capacity wall mid-migration takes a week.

Single-region today: everything runs in the North India region. Multi-region Kubernetes is not a thing you can buy here yet — plan your DR accordingly and keep backups of anything stateful.


Where to go next


Billing

Managed Kubernetes is billed per hour of running state, on the same pay-as-you-go model as VMs, invoiced in INR with GST.

  • Worker nodes are billed at their flavor's rate, per node, for as long as the node exists. Scaling a node pool down stops the meter for the removed nodes.
  • Persistent volumes are billed as block storage at the standard SSD rate, whether or not a pod is currently mounting them. Deleting a PersistentVolumeClaim whose reclaim policy is Retain does not stop that meter — delete the volume too.
  • Load balancers created for a Service of type LoadBalancer are billed as load balancers, one per service, at the standard L4/L7 rate. Consolidate behind a single ingress if you're publishing many services.
  • Egress is free. Traffic leaving the cluster to the internet is not metered.

If you hit an allocation limit when creating a cluster or scaling a pool, email info@thewahda.com and we'll raise it.