Skip to content

Kubernetes on infrastructure you own

Built to be left.

A production Kubernetes platform with no lock-in by design. It runs on infrastructure you own, and if you walk away the cluster keeps running.

Bare metal usually costs far less than cloud VMs for the same workload. Most teams skip it because the upkeep is where they get hurt. Fynrig takes on that upkeep.

1 · Describe it

# cluster.yaml
kind: Cluster
metadata:
  name: prod
spec:
  provider: bmaas
  privateNetwork: true
  controlPlane:
    count: 3
    flavor: dedicated
  workerPools:
    - name: general
      count: 3
      flavor: dedicated

2 · Apply it

$ kubectl apply -f cluster.yaml
cluster.fynrig.com/prod created

# wipe → image → bootstrap
# → network → storage, on servers
# in your own account

3 · Use it

$ kubectl get nodes
NAME   STATUS  ROLES          AGE
cp-1   Ready   control-plane  9m
cp-2   Ready   control-plane  8m
cp-3   Ready   control-plane  8m
w-1    Ready   worker         6m
w-2    Ready   worker         6m
w-3    Ready   worker         6m
Illustrative example of the workflow. Fynrig is operated from its dashboard and API.

Day 2 is where self-built clusters break

Provisioning is close to commoditized. Research across the free tooling in this space found the same four failures recurring, and Fynrig is built around each of them.

Built to catch a certificate before it expires

Fynrig is designed to track every expiry date from outside the cluster and rotate each certificate before the control plane stops trusting itself.

Built to catch a storage upgrade before it corrupts a volume

Storage volumes have degraded across version boundaries in production clusters. Fynrig is designed to sequence upgrades around volume health and hold one back when a volume is unhealthy.

Built to catch a drive dying before it takes a node with it

A node-level heartbeat is designed to run outside Kubernetes and report disk and liveness state even when the cluster is down, so a failing drive shows up before the node goes.

Built to catch the gap in your backups

An etcd snapshot cannot rebuild a cluster with persistent volumes attached. Fynrig is designed to back up cluster state and volumes together, so a restore brings the whole cluster back.

Standard Kubernetes on a standard base image.

Bare metal first

Built for rented dedicated servers, where tools like Metal3 and Tinkerbell cannot run because the provider exposes no BMC access. Fynrig wipes, images and bootstraps each node through the provider's rescue mode.

Any infrastructure provider

The provisioning layer is built to work across providers, so your cluster can follow the best price and hardware.

Nothing proprietary underneath

The node image is planned as open source, and the cluster is plain Kubernetes. The control plane runs in your account and the OS layer is a standard image.