Exercise 3: HPA Wants 20 Replicas, Only 12 More Fit — What Happens — Possible Solution ==================================================================== What happens to the remaining pods: HPA will still attempt to create all the pods needed to reach its target of 20 replicas -- 15 new pods beyond the original 5. Per this chapter's own "Full Autoscaling Picture" section, when the scheduler tries to place these new pods but existing nodes lack sufficient capacity, those pods that CAN'T be scheduled remain in the Pending phase -- directly per `k8s1-3`'s own pod lifecycle material ("Pending" meaning accepted by the API server but not yet fully scheduled/running). In this scenario, 12 of the 15 new pods would successfully find room and become Running, while the remaining 3 stay stuck in Pending, unable to be scheduled anywhere due to insufficient node capacity. What needs to happen next for all 20 to become Running: Per this chapter's own material, MORE NODE CAPACITY needs to become available -- specifically, Cluster Autoscaler needs to notice the Pending pods and provision additional node(s) to accommodate them. This is exactly the chapter's own described sequence: "HPA adds pods -> if no node has room, those pods stay Pending -> Cluster Autoscaler notices the Pending pods and adds a new node -> the pods get scheduled." Once Cluster Autoscaler adds enough new node capacity for the remaining 3 pods, the scheduler will place them onto the newly available nodes, and they'll transition from Pending to Running, finally reaching the full desired count of 20. If Cluster Autoscaler isn't configured or enabled at all, those 3 pods would remain stuck in Pending indefinitely -- HPA has no mechanism of its own to create new node capacity; it only manages pod COUNT within whatever capacity already exists (this chapter's own "HPA/VPA operate on pods within a fixed node set" distinction). WHY THIS WORKS AS AN ANSWER ------------------------------ This applies the chapter's own explicit multi-step scenario directly to the specific numbers given in the exercise (5 -> 20, but only 12 fit), correctly identifying that PARTIAL success is what happens (12 Running, 3 Pending) rather than either a full failure or a full success, and names Cluster Autoscaler specifically as the missing piece needed to resolve the remainder -- exactly the layered HPA-then-Cluster-Autoscaler relationship the chapter's own "Full Autoscaling Picture" section describes.