proper 1m_successful alarm for web_log

diff --git a/conf.d/health.d/web_log.conf b/conf.d/health.d/web_log.conf
index fa88dc8cf55aef26faff5eab5bd578ee1a753694..ad9e0161035375d64701da69bb8226d5fafd483e 100644 (file)
--- a/conf.d/health.d/web_log.conf
+++ b/conf.d/health.d/web_log.conf
@@ -32,10 +32,19 @@ families: *
     calc: ($this == 0)?(1):($this)
    units: requests
    every: 10s
-    warn: ($1m_requests > 30) ? ($this > ($status >= $WARNING ) ? ( 1 ) : ( 2 )) : ( 0 )
-    crit: ($1m_requests > 30) ? ($this > ($status == $CRITICAL) ? ( 2 ) : ( 5 )) : ( 0 )
-   delay: down 15m multiplier 1.5 max 1h
     info: the sum of successful HTTP requests over the last minute
+
+template: 1m_successful
+      on: web_log.response_codes
+families: *
+    calc: $1m_2xx * 100 / $1m_requests
+   units: %
+   every: 10s
+    warn: ($1m_requests > 30) ? ($this < (($status >= $WARNING ) ? ( 98 ) : ( 95 )) ) : ( 0 )
+    crit: ($1m_requests > 30) ? ($this < (($status == $CRITICAL) ? ( 95 ) : ( 90 )) ) : ( 0 )
+   delay: down 15m multiplier 1.5 max 1h
+    info: the ratio of HTTP redirects (3xx) vs the successful requests, \
+          over the last minute
       to: webmaster
 
 template: 1m_redirects