RYMVARD

Platform engineering and Kubernetes

Every namespace counted.Every idle request found.

For platform teams and providers that run Kubernetes for many teams and customers, on their own hosts or in a cloud account: Rymvard shows capacity per cluster, node pool, team and namespace — requests, limits and use — and names the teams that hold capacity they never use.

Read-only. New in version 0.28 — the screens show an illustrative example estate, not a customer.

Requested is not used.

A namespace requests what it might need and keeps it. The scheduler counts the request, the node sits idle, and the next team is told the cluster is full.

Requests against use.

Per namespace: requested, limited and used cores, memory and GPUs — and the tenants requesting far above what they use.

Pools that cannot deliver.

Node pools whose namespace quotas promise more than their allocatable capacity can give.

Every namespace has an owner.

Tag rules map namespace labels to customers and teams; a namespace without one is counted and named.

What Rymvard shows

Every cluster, pool and namespace.

Clusters with their distribution and freshness, node pools with allocatable, requested, used and quota — read from the API server and kube-state-metrics.

Rymvard's Platform page: Kubernetes clusters and node pools with allocatable, requested, used and quota capacity, and the tenants holding capacity they do not use
Platform: what each pool can give, what was requested and what is used.

Clusters where they run.

Kubernetes on your own hosts beside the ledger's servers, or in a customer's cloud account — counted against the capacity underneath.

GPUs in the pool.

GPUs requested against GPUs allocatable per node pool, beside the GPU estate on the GPU pages.

Subscriptions under the platform.

Where OpenShift runs on your hosts, its worker nodes count in the Red Hat subscription position.

Unknown, not zero.

A cluster read longer ago than its freshness limit shows unknown — never an empty pool.

Read-only, from the clusters.

Rymvard reads requests, limits and use. It never changes a cluster.

  • The Kubernetes API server and kube-state-metrics, read-only.
  • Namespace labels and tag rules for customer and team.
  • A cluster older than its freshness limit is unknown — never empty.
  • Detailed records stay in their country; customer names are visible only to the people who need them.

New in 0.28 · Workspaces

A home for platform engineering.

Platform teams open Rymvard on their own home — tenants requesting far above use, requested but unused cores, pools that cannot deliver their quotas and namespaces without an owner, with AI that explains every number. See the workspaces

Rymvard's Platform Engineering home: what changed since the last visit, suggested next steps, the tenants holding capacity they do not use and the node pools against their quotas
Platform Engineering

Platform Engineering

Which teams have room — and which hold capacity they never use?

  • Tenants requesting far above use
  • Requested but unused cores
  • Pools with quotas they cannot deliver
  • Tenants without a customer or team

Screens from the running product on its example estate. Customer names appear only for planners and admins — findings never name a customer.

Across every estate

One customer, every spelling.

The customer registry resolves every name in the ledger — contracts, racks, cloud accounts, licences — to one customer with its aliases and business groups, and billing and energy statements count it once. Takeover without migration

Give the idle cores back.

Your clusters, read-only: every namespace with its owner, its request and its use.

Write to contact@rymvard.com