Description
Cloud Monitoring collects operational metrics such as CPU, memory, network and cluster health data to help identify problems. Disabling required collection can delay detection of performance degradation or resource shortages.
GKE defaults can enable collection when monitoringService is omitted, so check the effective configuration. Enabling collection alone does not create alerting policies or collect every application metric.
Potential impact
- Performance degradation or resource shortages can go unnoticed for longer.
- Alert-based response and capacity planning can lack necessary information.
Remediation
- Identify required metrics and check the cluster's collection settings, permissions and actual incoming data.
- Use a supported Cloud Monitoring configuration and configure alerting policies and recipients for key metrics. Test the notification path.
Examples
These partial GKE Cluster request bodies illustrate settings from historical Deployment Manager documentation, whose support has ended. Required cluster settings are omitted; these are not complete creation requests.
Before
yaml
name: my-cluster
description: cluster
After
yaml
name: my-cluster
description: cluster
monitoringService: monitoring.googleapis.com/kubernetes
Explanation:
- Before: The monitoring value is omitted. Check defaults and effective settings to determine whether collection is active.
- After: Cloud Monitoring for GKE is explicit. Verify metric delivery and alert configuration separately.