This topic describes the alert metrics that are supported by Kubernetes Monitoring. You can select an alert metric when you create an alert rule.

Alert metric Description Filtering dimension
CoreDNS failure rate The threshold can be customized based on your business requirements. None
CoreDNS cache hit rate The threshold can be customized based on your business requirements. None
Slow call rate A slow call meets the following requirements:
  • DNS response time ≥ 100 ms
  • Response time for other metrics ≥ 500 ms
  • Protocol:
    • HTTP
    • MySQL
    • Redis
    • Kafka
    • DNS
  • Namespace
  • Type:
    • Service
    • Deployment
    • DaemonSet
    • StatefulSet
  • Resource name
  • Request content:
    • For the HTTP protocol, URL paths are used as the filter conditions.
    • For the Redis protocol, commands are used as the filter conditions.
    • For the MySQL protocol, SQL statements are used as the filter conditions.
Number of errors This alert metric is applicable to health checks and business that is sensitive to the number of errors.
Error rate This alert metric is applicable to various scenarios.
Requests/QPS This alert metric is applicable to scenarios such as QPS monitoring and throttling.
P99 response time The threshold can be customized based on your business requirements.
P95 response time The threshold can be customized based on your business requirements.
P90 response time The threshold can be customized based on your business requirements.
P50 response time The threshold can be customized based on your business requirements.
Average response time The threshold can be customized based on your business requirements.