On-premises and colocation
Bare metal workers. A control plane you don’t have to run.
For teams running Kubernetes on their own servers, in their own data centers or in colocation. Your servers do the work. We run the control plane: API server, etcd, scheduler, and controllers.
The problem
1.
etcd is the hard part
Quorum, compaction, defragmentation, backups, and certificate rotation are where self-run clusters go wrong.
2.
Shared fate with the racks
When control plane members sit in the same racks as workers, one power or network fault takes both down.
3.
Scaling means more state
More nodes mean more objects, more watches, and more load on the API server. It has to grow with the fleet.
What Kestrel runs for you
A dedicated, highly available Kubernetes control plane. Your workers stay on your nodes.
Dedicated etcd, run for you
A dedicated etcd cluster with its own certificate authority and regular snapshots. You never touch quorum.
Highly available
Control planes run across failure zones, away from your racks.
Your provisioning stays
Workers join with standard Kubernetes bootstrap tooling, so however you install servers today keeps working.
Help with the hard parts
Architecture reviews, upgrade planning, and security reviews from people who have run large clusters.
Questions
Things people ask us.
Which servers can we use?
Any server that runs a Linux distribution supported by the kubelet. We don’t manage the hardware itself.
Do you install or manage our servers?
No. Hardware, operating systems, and provisioning stay with you. We run the control plane and can advise on the rest.
Is the control plane shared with other customers?
No. Every customer gets a dedicated API server, etcd cluster, and certificate authority.
Also built for