Add documentation on the counters to Readme

README.md

@@ -100,6 +100,43 @@ In both of the above instances, the Google Cloud Platform resources are
 deployed using Terraform. Please see the [Terraform instructions](terraform/README.md)
 for more information on the deployment options available.
 
+## Monitoring
+
+The autoscaler publishes the following metrics to Cloud Monitoring which can be used to monitor the behavior of the autoscaler, and to configure alerts.
+
+
+### Poller
+
+* Message processing counters:
+    * `spanner-autoscaler/poller/requests-success` - the number of polling request messages recieved and processed successfully.
+    * `spanner-autoscaler/poller/requests-success` - the number of polling request messages which failed processing.
+
+* Spanner Instance polling counters:
+   * `spanner-autoscaler/poller/polling-success` - the number of successful polls of the Spanner instance metrics.
+   * `spanner-autoscaler/poller/polling-failed` - the number of failed polls of the Spanner instance metrics.
+   * Both of these metrics have `projectid` and `instanceid` to identify the Spanner instance.
+
+### Scaler
+
+* Message processing counters:
+    * `spanner-autoscaler/scaler/requests-success` - the number of scaling request messages recieved and processed successfully.
+    * `spanner-autoscaler/scaler/requests-success` - the number of scaling request messages which failed processing.
+* Spanner Instance scaling counters:
+    * `spanner-autoscaler/scaler/scaling-success` - the number of succesful rescales of the Spanner instance.
+    * `spanner-autoscaler/scaler/scaling-denied` - the number of Spanner instance rescale attempts that failed
+    * `spanner-autoscaler/scaler/scaling-failed` - the number of Spanner instance rescale attempts that were denied by autoscaler configuration or policy.
+
+    * These threww metrics have the following attributes:
+      * `projectid` - Spanner Project ID
+      * `instanceid` - Spanner Instance ID
+      * `method` - the scaling method used
+      * `direction` - which can be `SCALE_UP`, `SCALE_DOWN` or `SCALE_SAME` (when the calculated rescale size is equal to the current size)
+    * In addition, the `scaling-denied` counter has a `reason` attribute containing the reason why the scaling was not performed, which can be:
+      * `SAME_SIZE` - when the calculated rescale size is equal to the current instance size.
+      * `MAX_SIZE` - when the instance has already been scaled up to the maximum configured size.
+      * `WITHIN_COOLDOWN` - when the instance has been recently rescaled, and the autoscaler is waiting for the cooldown period to end.
+
+
 ## Configuration
 
 The parameters for configuring the Autoscaler are identical regardless of the chosen