Set a minimum threshold for error rate alerts

This avoids very small numbers of errors on idle machines
triggering an errir rate alert.

diff --git a/cookbooks/prometheus/templates/default/alert_rules.yml.erb b/cookbooks/prometheus/templates/default/alert_rules.yml.erb
index dfb83891d33b18040190d67359e071e0fd29a78f..b78bbaddd51fb8a54cc2834c4815f28a45c85af4 100644 (file)
--- a/cookbooks/prometheus/templates/default/alert_rules.yml.erb
+++ b/cookbooks/prometheus/templates/default/alert_rules.yml.erb
@@ -798,7 +798,7 @@ groups:
   - name: web
     rules:
       - alert: web error rate
   - name: web
     rules:
       - alert: web error rate

-        expr: sum(rate(api_call_count_total{status=~"50[0-8]|5[1-9][0-9]"}[5m])) by (instance) / sum(rate(api_call_count_total[5m])) by (instance) > 0.002
+        expr: sum(rate(api_call_count_total{status=~"50[0-8]|5[1-9][0-9]"}[5m])) by (instance) / sum(rate(api_call_count_total[5m])) by (instance) > 0.002 and sum(rate(api_call_count_total{status=~"50[0-8]|5[1-9][0-9]"}[5m])) by (instance) > 0.01

         for: 5m
         labels:
           alertgroup: web
         for: 5m
         labels:
           alertgroup: web