use quantile instead of average for error codes

files/core_alerts.yml

@@ -124,7 +124,7 @@ groups:
       summary: Prometheus DNS Domain (instance {{ $labels.instance }} {{ $labels.domain }})
       description: "The Domain Check has failed. The DNS server may have crashed.\n VALUE = {{ $value }}\n  LABELS = {{ $labels }}"
   - alert: HTTP status code is not 200 or 204
-    expr: ((avg without(job)(probe_http_status_code)) != 200) and ((avg without(job)(probe_http_status_code)) != 204)
+    expr: ((quantile by (instance) (0.5, probe_http_status_code)) != 200) and ((quantile by (instance) (0.5, probe_http_status_code)) != 204)
     for: 2m
     labels:
       severity: critical