In this lesson, you will learn how to use ResourceQuotas with scopeSelector to prevent critical workloads from being starved by non-essential tasks in your cluster.
The Problem: Global Quota Blindness
By default, a ResourceQuota applied to a namespace limits the total consumption of that namespace. However, if you run both critical services (like your API gateway) and background jobs (like data processing) in the same namespace, the background jobs could potentially consume all available quota. This leads to resource starvation, where your critical pods fail to schedule because the namespace has hit its aggregate limit.
The Solution: Quota Scoping
"Quota Scoping" allows you to partition your ResourceQuota based on the PriorityClass of your pods. By defining a scopeSelector, you can create dedicated "buckets" of resources within a single namespace. You can guarantee a specific amount of CPU and memory for high-priority pods, ensuring they are never blocked by low-priority background processes.
Consider this configuration:
apiVersion: v1
kind: ResourceQuota
metadata:
name: critical-pods-quota
spec:
hard:
requests.cpu: "4"
requests.memory: 8Gi
scopeSelector:
matchExpressions:
- operator: In
scopeName: PriorityClass
values: ["cluster-critical"]
In this example, this quota only applies to pods with the cluster-critical PriorityClass. Because the scope is restricted, these pods have a dedicated resource pool that cannot be exhausted by pods assigned to a "best-effort" or "batch" PriorityClass. Even if the rest of your namespace is "full" of low-priority tasks, the scheduler will recognize that the cluster-critical bucket still has headroom, effectively preventing resource starvation for your most important workloads.
By explicitly scoping your quotas to match your PriorityClass hierarchy, you decouple the resource consumption of non-essential tasks from your production traffic. This ensures that your system maintains a reliable performance baseline regardless of what other workloads are running in the namespace.
Key Takeaway: Use scopeSelector within your ResourceQuota to carve out guaranteed resource tiers based on PriorityClass, ensuring critical workloads remain insulated from noisy neighbors.