Node fleet, mission 8 of 8
Checkout first
advanced, about 15 minutes, in the Kubernetes Simulator
Opens once you pass Not on spot.
The briefing
From Leo Martins, customer support at Larchwood (a fictional company)
Customers are writing in: "checkout spins forever", "is your site down?". It started when the nightly reports kicked off. My guess is a server crashed?
The reports (batch) have to finish today too, so please don't just switch them off. checkout.yaml and batch.yaml are in the working directory.
What you'll do
- checkout: 1 ready replica
- checkout runs at a higher priority than batch
- batch keeps its spec: 4 replicas asking for 5 CPUs each
- Bonus star: Delete no pod by hand: let the scheduler choose what gives way
The mission checks the cluster, not your keystrokes: any fix that works, passes.
Concepts
Priority and preemption
A PriorityClass gives pods a number;
Read the Kubernetes docspriorityClassNamein a pod spec picks one. When a pod can't fit, the scheduler may evict lower-priority pods on a node to make room — preemption. The evicted pods go back to Pending like any other. Two classes are built in for system pods.
More Node fleet missions
- 1. The right hardware
- 2. Reserved seating
- 3. It doesn't fit
- 4. Kernel patch
- 5. One basket
- 6. All on one
- 7. Not on spot
- 8. Checkout first