haproxy/meta/prometheus.yml

{%- if pillar.haproxy is defined and pillar.haproxy.proxy is defined %}
{%- from "haproxy/map.jinja" import proxy with context %}
{%- if proxy.enabled and proxy.listen is defined and proxy.listen|length > 0 %}
server:
  alert:
    HaproxyDown:
{% raw %}
      if: >-
        haproxy_up != 1
      labels:
        severity: warning
        service: haproxy
      annotations:
        summary: 'Haproxy service down'
        description: 'Haproxy service is down on node {{ $labels.host }}'
{% endraw %}
{%- for listen_name, listen in proxy.listen.iteritems() if listen.get('check', True) %}
{%- set camel_case_name = listen_name.replace('-','_').split('_')|map('capitalize')|join('') %}
    HAproxy{{ camel_case_name }}HTTPResponse5xx:
{% raw %}
      if: >-
        rate(haproxy_http_response_5xx{sv="FRONTEND",proxy="{% endraw %}{{ listen_name }}{% raw %}"}[1m]) > 1
      for: 2m
      labels:
        severity: warning
        service: "haproxy/{{ $labels.proxy }}"
      annotations:
        summary: HTTP 5xx responses on '{{ $labels.proxy }}' proxy (host {{ $labels.host }})
        description: >-
          Too many 5xx HTTP errors have been detected on the '{{ $labels.proxy }}' proxy for the last 2 minutes
          ({{ $value }} error(s) per second)
{% endraw %}
    HAproxy{{ camel_case_name }}BackendWarning:
{% raw %}
      if: >-
        max(max_over_time(haproxy_active_servers{sv="BACKEND",proxy="{% endraw %}{{ listen_name }}{% raw %}"}[12h])) by (proxy)
        - min(haproxy_active_servers{sv="BACKEND",proxy="{% endraw %}{{ listen_name }}{% raw %}"}) by (proxy) >= 1
      for: 5m
      labels:
        severity: warning
        service: "haproxy/{{ $labels.proxy }}"
      annotations:
        summary: "At least one backend is down for '{{ $labels.proxy }}' proxy for the last 5 minutes"
        description: >-
           {{ $value }} of backends are down for the '{{ $labels.proxy }}' proxy
{% endraw %}
    HAproxy{{ camel_case_name }}BackendCritical:
{% raw %}
      if: >-
        (max(max_over_time(haproxy_active_servers{sv="BACKEND",proxy="{% endraw %}{{ listen_name }}{% raw %}"}[12h])) by (proxy)
         - min (haproxy_active_servers{sv="BACKEND",proxy="{% endraw %}{{ listen_name }}{% raw %}"}) by (proxy)
        ) / max(max_over_time(haproxy_active_servers{sv="BACKEND",proxy="{% endraw %}{{ listen_name }}{% raw %}"}[12h])) by (proxy) * 100 >= 50
      for: 5m
      labels:
        severity: critical
        service: "haproxy/{{ $labels.proxy }}"
      annotations:
        summary: "Less than 50% of backends are up for the '{{ $labels.proxy }}' proxy for the last 5 minutes"
        description: >-
           {{ $value }}% of backends are down for the '{{ $labels.proxy }}' proxy
{% endraw %}
    HAproxy{{ camel_case_name }}BackendDown:
{% raw %}
      if: >-
        max(haproxy_active_servers{sv="BACKEND",proxy="{% endraw %}{{ listen_name }}{% raw %}"}) by (proxy)
        + max(haproxy_backup_servers{sv="BACKEND",proxy="{% endraw %}{{ listen_name }}{% raw %}"}) by (proxy) == 0
      for: 2m
      labels:
        severity: down
        service: "haproxy/{{ $labels.proxy }}"
      annotations:
        summary: "All backends are down for the '{{ $labels.proxy }}' proxy"
        description: >-
            The proxy '{{ $labels.proxy }}' has no active backend
{% endraw %}
{%- endfor %}
{%- endif %}
{%- endif %}