fbpx

Kubernetes Clusters

Each node in the cluster is monitored by the master nodes and assigned with a condition (state), that describes the status of different aspects of the node. The conditions types are: OutOfDisk, Ready, MemoryPressure, DiskPressure, NetworkUnavailable. Each condition type is assigned with a status: False, True or Unknown.
To ensure optimal performances, Ocean monitors the nodes’ status every 30 seconds and in case it identifies that the Ready condition is False or Unknown, it considers this node as Unhealthy and triggers a replacement.

 

ECS Clusters

Every Container Instance within the Amazon ECS cluster has the following two parameters assigned to it:

  1. status – The valid values are REGISTERING , REGISTRATION_FAILED , ACTIVE , INACTIVE , DEREGISTERING , or DRAINING
  2. agentConnected – This parameter returns true if the agent is connected to Amazon ECS. Registered instances with an agent that may be unhealthy or stopped return false.

To ensure cluster functionality and availability, Ocean monitors those parameters and will consider a Container Instance as Healthy when its status value is either ACTIVE or  DRAINING and the agentConnected value returns true .

When either of the above parameters return different values, or when the Container Instance does not respond at all, Ocean will consider the Container Instance Unhealthy and a replacement will be triggered.