octavia/meta/prometheus.yml

{%- from "octavia/map.jinja" import api, monitoring with context %}

{%- if api.get('enabled', False) %}
server:
  alert:
{%- raw %}
    OctaviaApiDown:
      if: >-
        max(openstack_api_check_status{service="octavia-api"}) == 0
      for: 2m
      labels:
        severity: critical
        service: octavia
      annotations:
        summary: "Octavia API outage"
        description: >-
          Octavia API is not accessible for all available Octavia endpoints in the OpenStack service catalog for 2 minutes.
{%- endraw %}
    OctaviaErrorLogsTooHigh:
      {%- set log_threshold = monitoring.error_log_rate|float %}
      if: >-
        sum(rate(log_messages{service="octavia",level=~"error|emergency|fatal"}[5m])) without (level) > {{ log_threshold }}
{%- raw %}
      labels:
        severity: warning
        service: octavia
      annotations:
        summary: "High number of errors in Octavia logs"
        description: "The average per-second rate of errors in Octavia logs on the {{ $labels.host }} node is {{ $value }} (as measured over the last 5 minutes)."
{%- endraw %}
{%- endif %}