#Controlplane alternative update
#Controlplane alternative upgrade

a MachineHealthCheck marks a machine for remediation. To support auto remediation triggered by MachineHealthCheck objects:Į.g.

Control Plane Provider reconciles by removing the corresponding etcd member and updating related metadata To manage control plane deployments across failure domains.Į.g.To manage a kubeadm-based, 'external etcd' control plane (using a pre-existing, user-managed, etcd clusters).To manage a kubeadm-based, 'stacked etcd' control plane.

#Controlplane alternative update

To support Rolling Update type of rollout strategy (similar to MachineDeployment) in KubeadmControlPlane.To enable declarative orchestrated replacement of control plane machines, such as to roll out an OS-level CVE fix.To provide a kubeadm-based implementation that is infrastructure provider agnostic.To provide a default machine-based implementation using kubeadm.To enable declarative orchestrated control plane upgrades.To enable scaling of the number of control plane nodes.To support single node and multiple node control plane instances, with the requirement that the infrastructure provider supports some type of a stable endpoint for the API Server (Load Balancer, VIP, etc).To establish new resource types for control plane management.We also wanted to ensure that default control plane management and use of any alternative control plane management solutions are separated. Much likeīootstrapping was identified as being reimplemented in every infrastructure provider and then extracted into Cluster APIīootstrap Provider Kubeadm (CABPK), we believe we can reduce the redundancy of control plane management across providersĪnd centralize the logic in Cluster API. Motivationĭuring 2019 we saw control plane management implementations in each infrastructure provider. The control plane covered by this document is defined as the Kubernetes API server, scheduler, controller manager, DNSĪnd proxy services, and the underlying etcd data store. Upgrading, scaling up, and modifying the underlying image (e.g. This proposal outlines a new process for Cluster API to manage control plane machines as a single concept. Kubernetes Control Plane Management Overview Summary Any discrepancies should be rectified in the main Cluster API glossary. The lexicon used in this document is described in more detail here.

#Controlplane alternative upgrade

Upgrade where changes needed to KubeadmConfig are not currently possible.

Adoption of pre-v1alpha3 Control Plane Machines.

Scenario 4: Unhealthy machines combined with rollout.

Scenario 3: Three replicas, one unresponsive etcd member, one (different) unhealthy machine.

Scenario 2: Three replicas, two machines marked for remediation.

Scenario 1: Three replicas, one machine marked for remediation.

Remediation (using delete-and-recreate).

Delete of the entire KubeadmControlPlane (kubectl delete controlplane my-controlplane).

Modifications required to existing API Types.

Implementation Details/Notes/Constraints.

Kubeadm Based Control Plane Management Table of Contents