Which capability ensures Kubernetes automatically restarts failed containers and reschedules them on available nodes?

Which capability ensures Kubernetes automatically restarts failed containers and reschedules them on available nodes?

• A. Auto-scaling
• B. Rolling updates
• C. Blue-green deployment
• D. Self-healing

Answer: (D)

Self-healing is the capability that ensures Kubernetes automatically restarts failed containers and reschedules them on available nodes. Kubernetes maintains the desired state of your applications; when a container crashes or becomes unhealthy, the control plane detects the issue and creates a new pod on a healthy node, keeping the service up. Health checks like liveness probes trigger restarts when needed, and the scheduler places the replacement pod on a node with capacity, so the workload remains available even if something goes wrong on a node. Auto-scaling changes how many pods run based on load, not automatically restarting failed containers. Rolling updates are about progressively updating versions with minimal downtime, not fault recovery. Blue-green deployment is a release strategy using two parallel environments and traffic switching, not automatic recovery of failed components.

Self-healing is the capability that ensures Kubernetes automatically restarts failed containers and reschedules them on available nodes. Kubernetes maintains the desired state of your applications; when a container crashes or becomes unhealthy, the control plane detects the issue and creates a new pod on a healthy node, keeping the service up. Health checks like liveness probes trigger restarts when needed, and the scheduler places the replacement pod on a node with capacity, so the workload remains available even if something goes wrong on a node.

Auto-scaling changes how many pods run based on load, not automatically restarting failed containers. Rolling updates are about progressively updating versions with minimal downtime, not fault recovery. Blue-green deployment is a release strategy using two parallel environments and traffic switching, not automatic recovery of failed components.