From d5c5e2264043830c242ee975b0dbb2950e508523 Mon Sep 17 00:00:00 2001
From: ale <ale@incal.net>
Date: Mon, 10 Jun 2019 19:05:50 +0100
Subject: [PATCH] NGINX error metrics should only include 5xx errors

Previously we were also counting 4xx errors, which are not relevant
as a reliability indicator. We may eventually want to restrict this
further to 502/503 (for backend timeouts only).
---
 roles/prometheus/files/rules/alerts_nginx.conf.yml | 4 ++--
 roles/prometheus/files/rules/rules_nginx.conf.yml  | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/roles/prometheus/files/rules/alerts_nginx.conf.yml b/roles/prometheus/files/rules/alerts_nginx.conf.yml
index 2b5d386..790a122 100644
--- a/roles/prometheus/files/rules/alerts_nginx.conf.yml
+++ b/roles/prometheus/files/rules/alerts_nginx.conf.yml
@@ -10,7 +10,7 @@ groups:
       severity: page
     annotations:
       summary: 'High HTTP error ratio for {{$labels.vhost}} globally'
-      description: 'We are serving lots of 4xx/5xx errors for {{$labels.vhost}} on all frontends.'
+      description: 'We are serving lots of 5xx errors for {{$labels.vhost}} on all frontends.'
   - alert: HTTPErrorRatioHigh
     expr: (instance:nginx_http_requests_errs:ratio > 0.02 and instance:nginx_http_requests_total:rate5m > 0.1)
     for: 10m
@@ -20,4 +20,4 @@ groups:
       severity: page
     annotations:
       summary: 'High HTTP error ratio for {{$labels.vhost}} on {{$labels.host}}'
-      description: 'We are serving lots of 4xx/5xx errors for {{$labels.vhost}} on {{$labels.host}}.'
+      description: 'We are serving lots of 5xx errors for {{$labels.vhost}} on {{$labels.host}}.'
diff --git a/roles/prometheus/files/rules/rules_nginx.conf.yml b/roles/prometheus/files/rules/rules_nginx.conf.yml
index a37d670..1714c41 100644
--- a/roles/prometheus/files/rules/rules_nginx.conf.yml
+++ b/roles/prometheus/files/rules/rules_nginx.conf.yml
@@ -4,14 +4,14 @@ groups:
   - record: instance:nginx_http_requests_total:rate5m
     expr: sum(rate(nginx_http_requests[5m])) without (backend, method, code)
   - record: instance:nginx_http_requests_errs:rate5m
-    expr: sum(rate(nginx_http_requests{code=~"[45].*"}[5m])) without (backend, method, code)
+    expr: sum(rate(nginx_http_requests{code=~"5.*"}[5m])) without (backend, method, code)
   - record: instance:nginx_http_requests_errs:ratio
     expr: (instance:nginx_http_requests_errs:rate5m / instance:nginx_http_requests_total:rate5m)
   - record: global:nginx_http_requests_total:rate5m
     expr: sum(rate(nginx_http_requests[5m])) without (instance, host, exported_instance,
       exported_host, backend, method, code)
   - record: global:nginx_http_requests_errs:rate5m
-    expr: sum(rate(nginx_http_requests{code=~"[45].*"}[5m])) without (instance, host,
+    expr: sum(rate(nginx_http_requests{code=~"5.*"}[5m])) without (instance, host,
       exported_instance, exported_host, backend, method, code)
   - record: global:nginx_http_requests_errs:ratio
     expr: (global:nginx_http_requests_errs:rate5m / global:nginx_http_requests_total:rate5m)
-- 
GitLab