##告警配置项
sum(dcgm_gpu_utilization{job="gpu-exporter"}) by (instance,gpu) > 80

labels变量保存警报实例的标签键/值对,value保存警报实例的评估值
###报警规则文件
prometheus/prometheus.yml
rule_files:
- "rules/*.yml"
检测规则正确性
./promtool check rules rules/blackbox_port_check.yml
Checking rules/blackbox_port_check.yml
SUCCESS: 3 rules found