Fix ResourceQuota Exceeded Errors

💡 Quick Answer: forbidden: exceeded quota means the namespace has a ResourceQuota and your pod’s requests exceed the remaining budget. Check kubectl describe quota -n <ns> to see used vs. hard limits. Free up quota by deleting unused pods or increasing the quota.

The Problem

Pod creation fails with:

Error from server (Forbidden): pods "myapp-xyz" is forbidden:
  exceeded quota: compute-quota, requested: cpu=500m,memory=512Mi,
  used: cpu=3500m,memory=7Gi, limited: cpu=4,memory=8Gi

The Solution

Step 1: Check Current Quota Usage

kubectl describe quota -n myapp
# Name:           compute-quota
# Resource        Used    Hard
# --------        ----    ----
# cpu             3500m   4       ← Only 500m remaining
# memory          7Gi     8Gi     ← Only 1Gi remaining
# pods            7       20
# requests.cpu    3500m   4
# requests.memory 7Gi     8Gi

Step 2: Find What’s Consuming Quota

# List all pods and their resource requests
kubectl get pods -n myapp -o json | jq -r '
  .items[] |
  .metadata.name as $name |
  .spec.containers[] |
  "\($name): cpu=\(.resources.requests.cpu // "none") memory=\(.resources.requests.memory // "none")"
'

Step 3: Fix — Free Up or Increase Quota

# Option A: Delete unused pods/deployments
kubectl delete deploy old-service -n myapp

# Option B: Reduce resource requests
kubectl set resources deploy myapp --requests=cpu=200m,memory=256Mi -n myapp

# Option C: Increase quota (requires cluster-admin)
kubectl patch resourcequota compute-quota -n myapp --type merge -p '{
  "spec": {"hard": {"cpu": "8", "memory": "16Gi"}}
}'

LimitRange Defaults

If pods don’t specify resources, a LimitRange may inject defaults that count against quota:

kubectl describe limitrange -n myapp
# Default Request: cpu=250m, memory=256Mi
# Even pods WITHOUT explicit requests get these defaults → they consume quota

Common Issues

Quota Counts Terminating Pods

Pods in Terminating state still count. Force-delete stuck terminating pods to free quota.

Quota Blocks Scaling

HPA can’t scale up because quota is exhausted. Either increase quota or set conservative HPA maxReplicas.

Best Practices

• Always set resource requests — without them, LimitRange defaults apply (which may be too high)
• Monitor quota usage — alert at 80% to prevent surprise failures
• Use separate quotas per team — prevents one team from consuming all resources
• Set both requests and limits quotas — requests.cpu for scheduling, limits.cpu for burst

Key Takeaways

• ResourceQuota caps total resource requests per namespace
• kubectl describe quota shows used vs hard limits at a glance
• LimitRange injects default requests — these count against quota even if you didn’t set them
• Free quota by deleting pods, reducing requests, or increasing the quota