CPU and memory limits are set

Each instance declares how much CPU and memory it needs and may use.

MediumChecked automatically

Applies to:
Services on platforms where you set replicas and resource limits
Rule id:
deployment.resource-limits

Why it matters

Without a declared request the scheduler packs the service onto a node by guesswork, and a noisy neighbor starves it. Without a limit a memory leak takes the node down along with everything else on it. Stated numbers are also the starting point for capacity planning and for cost.

How to apply it

  • Set a memory request and limit from what the service uses under load, with headroom. Keep them equal or close, so the instance is not the first one evicted.
  • Set a CPU request from measured use. Be careful with tight CPU limits: throttling shows up as latency, not as an error.
  • Revisit the numbers after load tests and after traffic grows.

Example

deployment.resources: { cpu: 500m, memory: 512Mi }, measured at 300m and 380Mi at peak.

References

Checklists with this rule

Related agent skill: kubernetes-deployment