Add documentation for activator <> autoscaler connectivity metrics

docs/versioned/serving/observability/metrics/serving-metrics.md

@@ -92,6 +92,45 @@ Name | Type | Description
 `kn.configuration.name` | string | Knative Configuration name associated with this Revision
 `kn.revision.name` | string | The name of the Revision
 
+### `kn.activator.stats.conn.reachable`
+
+**Instrument Type:** Int64Gauge
+
+**Unit ([UCUM](https://ucum.org)):** {reachable}
+
+**Description:** Whether a peer is reachable from the activator (1 = reachable, 0 = not reachable)
+
+The following attributes are included with the metric
+
+Name | Type | Description
+-|-|-
+`peer` | string | The peer service the activator is connecting to (e.g., `autoscaler`)
+
+This metric helps operators identify connectivity issues between the activator and its peer components. The metric is recorded:
+
+- When a connection is established (value = 1)
+- When a connection is lost (value = 0)
+
+### `kn.activator.stats.conn.errors`
+
+**Instrument Type:** Int64Counter
+
+**Unit ([UCUM](https://ucum.org)):** {error}
+
+**Description:** Number of connection errors from the activator
+
+The following attributes are included with the metric
+
+Name | Type | Description
+-|-|-
+`peer` | string | The peer service the activator is connecting to (e.g., `autoscaler`)
+
+This counter increments each time the activator fails to communicate with a peer. It complements the `kn.activator.stats.conn.reachable` gauge by providing a cumulative count of errors, which is useful for:
+
+- Detecting flaky connections that might be missed by point-in-time gauge sampling
+- Creating rate-based alerts (e.g., alert if error rate exceeds threshold over 5 minutes)
+- Tracking connection stability trends over time
+
 ### HTTP metrics
 
 Since the activator receives and forwards requests to the user workload it has both HTTP server and client metrics.