Quiz 3 of 5: Scheduling, workloads and platforms
7
Three replicas of a service all ran on one node. The node failed and the service went down, although kubectl showed 3/3 Running earlier. What is the best fix?
Why: Replica count does not control placement. Spread constraints force an even distribution across failure domains, and a PDB protects the service during voluntary disruptions such as drains and upgrades.
8
A Deployment shows 2 of 3 replicas, and kubectl get pods lists no Pending Pod. What is the likely cause?
Why: Quota is enforced at admission, before scheduling. The Pod never exists, so there is nothing to be Pending. The evidence is in the ReplicaSet's events and kubectl describe resourcequota.
9
A Deployment with 3 replicas uses maxSurge 1 and maxUnavailable 0. What happens during a rolling update?
Why: maxSurge lets the controller create an extra Pod, and maxUnavailable 0 forbids dropping below the desired count. Readiness gates each step. Setting both to 0 is the combination that would deadlock.