MCS’s Health Checks and Auto Healing

MCS can ensure that the running Kubernetes worker nodes are healthy and that they are ready to serve your different Pods and application’s needs with MCS Auto Healing for Kubernetes.

With the MCS Auto Healing configured, MCS monitors the nodes’ status every 30 seconds, searching for unhealthy instances and triggering replacements when needed, maximizing your cluster’s efficiency and performance.

How Auto Healing Works

The MCS Auto Healing checks for the node’s status. The status of each Kubernetes node is represented as a “condition” object, that describes the status of different aspects of the node. The Auto Healing inspects the Ready condition. If the Ready condition has either the False or Unknown status, the instance is considered unhealthy and a replacement is triggered.

Configuring Auto Healing
  • Health Check Grace Period: The time to wait after the instance is launched before the first health check. This is the expected “warm up” period for the instance.
  • Unhealthy Duration: The Ready condition may be temporarily False or Unknown for a wide variety of reasons. The Unhealthy Duration period is the amount of time that the instance is found to be unhealthy before a replacement is triggered, ensuring that instances aren’t unnecessarily replaced. Since the health checks occur every 30 seconds, the Unhealthy Duration should be a minimum of 60 seconds.
In the Spotinst Console

In the Compute tab, under Load Balancers, select Kubernetes in the Auto Healing section.