Prometheus fire every alert when value changed

[ghost]

In prometheus , my cpu usage rule is like below:

- alert: cpu_usage
  annotations:
    description: ({{ $labels.instance }})
    summary: cpu_usage{{ $value }}%
  expr: |
    100 - floor((avg by(instance) (rate(node_cpu_seconds_total{service="node-exporter",mode="idle"}[2m])) * 100)) > 80
  for: 0m
  labels:
    servicee: node
    severity: warn
    value: '{{ $value }}'

I put the autual value in the lables and duration set to 0m(for: 0m). Current issue is: It will fire every alert after the value is beyond threadhold 80, for example, when it reaches 81, it will definately accord with the rule and fire the alert, howevery, after it comes to 82, there will be another alert of the same alertname "cpu_usage". repeat_interval seems doesn't work in alertmanage configuration.