Add a howto on alerting when % thresholds breached

howto.html

@@ -476,6 +476,39 @@ <h3>Find the host using the most CPU</h3>
   (with {:service "Max CPU" :host nil} prn)))
 {% endhighlight %}
 
+
+<h3>Alerting when a certain percentage of events happen</h3>
+
+<p>Sometimes you'll have a service that will fail. You might expect one or two
+failures, but if you get over a certain percentage of failures you want to be
+notified. In this case you can use a <code>fixed-time-window</code>
+
+{% highlight clj %}
+(streams
+  (where (and metric (service "app1")(tagged "sign_in")
+    ; We want to get alerted about failed sign ins. However we expect that there will be failures
+    ; due to incorect passwords etc. So we only want to get alerted if more than 50% of the signins
+    ; in a 60 second period are failures. The app tags failed signins with a warning state.
+    ;
+    ; fixed-time-window sends a vector of events out every 60 seconds
+    (fixed-time-window 60
+      ; smap passes those events into a function
+      (smap (fn [events]
+        ; we include a function here filter the vector for all of the "warning" events and count them
+        ; we will then count the total of the events
+        ; finally work out the percentage
+        (let [percent (/ (count (filter #(= (:state %) "warning") events))
+                         (count events))]
+                         ; take an action on the value of percent, in this example we imagine that the
+                         ; page-ops and mail-devs functions can take strings
+                         (cond
+                           (> percent 0.7) (page-ops (format "sign_in is CRITICAL: %f percent" (float percent)))
+                           (> percent 0.5) (mail-devs (format "sign_in is BAD: %f percent" (float percent)))
+                           :else (prn (format "sign in is %f" (float percent)))
+                           ))))))
+  ))
+{% endhighlight %}
+
 <h2>Working with dashboard</h2>
 
 <h3>Application specific host grouping</h3>