Metrics Server exposes core Kubernetes metrics via metrics API.
More details can be found in Core metrics pipeline documentation.
Metrics Server supports up to 30 pods per cluster node. In clusters where there are more running pods, Metrics Server may be throttled or fail with OOM error. Starting with Kubernetes 1.9.2, Metrics Server resource requirements may be overwritten manually. Learn more about Addon Resizer configuration
Decreasing resource requirements for cluster addons may cause system instability. The effects may include (but are not limited to):
kubectl top not working (starting with Kubernetes 1.10)Overwritten configuration persists through cluster updates, therefore may cause all effects above after a cluster update.