feat: expose current agent connections by type via prometheus (coder#14612)

ethanndickson · web-flow · commit c8580a415a45 · 2024-09-11T14:13:30.000+10:00
diff --git a/agent/agent.go b/agent/agent.go
@@ -1510,6 +1510,8 @@ func (a *agent) Collect(ctx context.Context, networkStats map[netlogtype.Connect
 	var mu sync.Mutex
 	status := a.network.Status()
 	durations := []float64{}
+	p2pConns := 0
+	derpConns := 0
 	pingCtx, cancelFunc := context.WithTimeout(ctx, 5*time.Second)
 	defer cancelFunc()
 	for nodeID, peer := range status.Peer {
@@ -1526,13 +1528,18 @@ func (a *agent) Collect(ctx context.Context, networkStats map[netlogtype.Connect
 		wg.Add(1)
 		go func() {
 			defer wg.Done()
-			duration, _, _, err := a.network.Ping(pingCtx, addresses[0].Addr())
+			duration, p2p, _, err := a.network.Ping(pingCtx, addresses[0].Addr())
 			if err != nil {
 				return
 			}
 			mu.Lock()
 			defer mu.Unlock()
 			durations = append(durations, float64(duration.Microseconds()))
+			if p2p {
+				p2pConns++
+			} else {
+				derpConns++
+			}
 		}()
 	}
 	wg.Wait()
@@ -1552,6 +1559,9 @@ func (a *agent) Collect(ctx context.Context, networkStats map[netlogtype.Connect
 	// Agent metrics are changing all the time, so there is no need to perform
 	// reflect.DeepEqual to see if stats should be transferred.
 
+	// currentConnections behaves like a hypothetical `GaugeFuncVec` and is only set at collection time.
+	a.metrics.currentConnections.WithLabelValues("p2p").Set(float64(p2pConns))
+	a.metrics.currentConnections.WithLabelValues("derp").Set(float64(derpConns))
 	metricsCtx, cancelFunc := context.WithTimeout(ctx, 5*time.Second)
 	defer cancelFunc()
 	a.logger.Debug(ctx, "collecting agent metrics for stats")
diff --git a/agent/agent_test.go b/agent/agent_test.go
@@ -2531,17 +2531,17 @@ func TestAgent_Metrics_SSH(t *testing.T) {
 	err = session.Shell()
 	require.NoError(t, err)
 
-	expected := []agentsdk.AgentMetric{
+	expected := []*proto.Stats_Metric{
 		{
 			Name:  "agent_reconnecting_pty_connections_total",
-			Type:  agentsdk.AgentMetricTypeCounter,
+			Type:  proto.Stats_Metric_COUNTER,
 			Value: 0,
 		},
 		{
 			Name:  "agent_sessions_total",
-			Type:  agentsdk.AgentMetricTypeCounter,
+			Type:  proto.Stats_Metric_COUNTER,
 			Value: 1,
-			Labels: []agentsdk.AgentMetricLabel{
+			Labels: []*proto.Stats_Metric_Label{
 				{
 					Name:  "magic_type",
 					Value: "ssh",
@@ -2554,30 +2554,46 @@ func TestAgent_Metrics_SSH(t *testing.T) {
 		},
 		{
 			Name:  "agent_ssh_server_failed_connections_total",
-			Type:  agentsdk.AgentMetricTypeCounter,
+			Type:  proto.Stats_Metric_COUNTER,
 			Value: 0,
 		},
 		{
 			Name:  "agent_ssh_server_sftp_connections_total",
-			Type:  agentsdk.AgentMetricTypeCounter,
+			Type:  proto.Stats_Metric_COUNTER,
 			Value: 0,
 		},
 		{
 			Name:  "agent_ssh_server_sftp_server_errors_total",
-			Type:  agentsdk.AgentMetricTypeCounter,
+			Type:  proto.Stats_Metric_COUNTER,
 			Value: 0,
 		},
 		{
-			Name:  "coderd_agentstats_startup_script_seconds",
-			Type:  agentsdk.AgentMetricTypeGauge,
+			Name:  "coderd_agentstats_currently_reachable_peers",
+			Type:  proto.Stats_Metric_GAUGE,
 			Value: 0,
-			Labels: []agentsdk.AgentMetricLabel{
+			Labels: []*proto.Stats_Metric_Label{
+				{
+					Name:  "connection_type",
+					Value: "derp",
+				},
+			},
+		},
+		{
+			Name:  "coderd_agentstats_currently_reachable_peers",
+			Type:  proto.Stats_Metric_GAUGE,
+			Value: 1,
+			Labels: []*proto.Stats_Metric_Label{
 				{
-					Name:  "success",
-					Value: "true",
+					Name:  "connection_type",
+					Value: "p2p",
 				},
 			},
 		},
+		{
+			Name:  "coderd_agentstats_startup_script_seconds",
+			Type:  proto.Stats_Metric_GAUGE,
+			Value: 1,
+		},
 	}
 
 	var actual []*promgo.MetricFamily
@@ -2586,17 +2602,33 @@ func TestAgent_Metrics_SSH(t *testing.T) {
 		if err != nil {
 			return false
 		}
-
-		if len(expected) != len(actual) {
-			return false
+		count := 0
+		for _, m := range actual {
+			count += len(m.GetMetric())
 		}
-
-		return verifyCollectedMetrics(t, expected, actual)
+		return count == len(expected)
 	}, testutil.WaitLong, testutil.IntervalFast)
 
-	require.Len(t, actual, len(expected))
-	collected := verifyCollectedMetrics(t, expected, actual)
-	require.True(t, collected, "expected metrics were not collected")
+	i := 0
+	for _, mf := range actual {
+		for _, m := range mf.GetMetric() {
+			assert.Equal(t, expected[i].Name, mf.GetName())
+			assert.Equal(t, expected[i].Type.String(), mf.GetType().String())
+			// Value is max expected
+			if expected[i].Type == proto.Stats_Metric_GAUGE {
+				assert.GreaterOrEqualf(t, expected[i].Value, m.GetGauge().GetValue(), "expected %s to be greater than or equal to %f, got %f", expected[i].Name, expected[i].Value, m.GetGauge().GetValue())
+			} else if expected[i].Type == proto.Stats_Metric_COUNTER {
+				assert.GreaterOrEqualf(t, expected[i].Value, m.GetCounter().GetValue(), "expected %s to be greater than or equal to %f, got %f", expected[i].Name, expected[i].Value, m.GetCounter().GetValue())
+			}
+			for j, lbl := range expected[i].Labels {
+				assert.Equal(t, m.GetLabel()[j], &promgo.LabelPair{
+					Name:  &lbl.Name,
+					Value: &lbl.Value,
+				})
+			}
+			i++
+		}
+	}
 
 	_ = stdin.Close()
 	err = session.Wait()
@@ -2828,28 +2860,6 @@ func TestAgent_ManageProcessPriority(t *testing.T) {
 	})
 }
 
-func verifyCollectedMetrics(t *testing.T, expected []agentsdk.AgentMetric, actual []*promgo.MetricFamily) bool {
-	t.Helper()
-
-	for i, e := range expected {
-		assert.Equal(t, e.Name, actual[i].GetName())
-		assert.Equal(t, string(e.Type), strings.ToLower(actual[i].GetType().String()))
-
-		for _, m := range actual[i].GetMetric() {
-			assert.Equal(t, e.Value, m.Counter.GetValue())
-
-			if len(m.GetLabel()) > 0 {
-				for j, lbl := range m.GetLabel() {
-					assert.Equal(t, e.Labels[j].Name, lbl.GetName())
-					assert.Equal(t, e.Labels[j].Value, lbl.GetValue())
-				}
-			}
-			m.GetLabel()
-		}
-	}
-	return true
-}
-
 type syncWriter struct {
 	mu sync.Mutex
 	w  io.Writer
diff --git a/agent/metrics.go b/agent/metrics.go
@@ -19,6 +19,7 @@ type agentMetrics struct {
 	// startupScriptSeconds is the time in seconds that the start script(s)
 	// took to run. This is reported once per agent.
 	startupScriptSeconds *prometheus.GaugeVec
+	currentConnections   *prometheus.GaugeVec
 }
 
 func newAgentMetrics(registerer prometheus.Registerer) *agentMetrics {
@@ -45,10 +46,19 @@ func newAgentMetrics(registerer prometheus.Registerer) *agentMetrics {
 	}, []string{"success"})
 	registerer.MustRegister(startupScriptSeconds)
 
+	currentConnections := prometheus.NewGaugeVec(prometheus.GaugeOpts{
+		Namespace: "coderd",
+		Subsystem: "agentstats",
+		Name:      "currently_reachable_peers",
+		Help:      "The number of peers (e.g. clients) that are currently reachable over the encrypted network.",
+	}, []string{"connection_type"})
+	registerer.MustRegister(currentConnections)
+
 	return &agentMetrics{
 		connectionsTotal:      connectionsTotal,
 		reconnectingPTYErrors: reconnectingPTYErrors,
 		startupScriptSeconds:  startupScriptSeconds,
+		currentConnections:    currentConnections,
 	}
 }
 
diff --git a/docs/admin/prometheus.md b/docs/admin/prometheus.md
@@ -96,6 +96,11 @@ spec:
 
 ## Available metrics
 
+`coderd_agentstats_*` metrics must first be enabled with the flag
+`--prometheus-collect-agent-stats`, or the environment variable
+`CODER_PROMETHEUS_COLLECT_AGENT_STATS` before they can be retrieved from the
+deployment. They will always be available from the agent.
+
 <!-- Code generated by 'make docs/admin/prometheus.md'. DO NOT EDIT -->
 
 | Name                                                          | Type      | Description                                                                                                                      | Labels                                                                              |
@@ -107,6 +112,7 @@ spec:
 | `coderd_agents_up`                                            | gauge     | The number of active agents per workspace.                                                                                       | `template_name` `username` `workspace_name`                                         |
 | `coderd_agentstats_connection_count`                          | gauge     | The number of established connections by agent                                                                                   | `agent_name` `username` `workspace_name`                                            |
 | `coderd_agentstats_connection_median_latency_seconds`         | gauge     | The median agent connection latency                                                                                              | `agent_name` `username` `workspace_name`                                            |
+| `coderd_agentstats_currently_reachable_peers`                 | gauge     | The number of peers (e.g. clients) that are currently reachable over the encrypted network.                                      | `agent_name` `connection_type` `template_name` `username` `workspace_name`          |
 | `coderd_agentstats_rx_bytes`                                  | gauge     | Agent Rx bytes                                                                                                                   | `agent_name` `username` `workspace_name`                                            |
 | `coderd_agentstats_session_count_jetbrains`                   | gauge     | The number of session established by JetBrains                                                                                   | `agent_name` `username` `workspace_name`                                            |
 | `coderd_agentstats_session_count_reconnecting_pty`            | gauge     | The number of session established by reconnecting PTY                                                                            | `agent_name` `username` `workspace_name`                                            |
diff --git a/scripts/metricsdocgen/metrics b/scripts/metricsdocgen/metrics
@@ -63,6 +63,9 @@ coderd_agentstats_connection_count{agent_name="main",username="admin",workspace_
 # HELP coderd_agentstats_connection_median_latency_seconds The median agent connection latency
 # TYPE coderd_agentstats_connection_median_latency_seconds gauge
 coderd_agentstats_connection_median_latency_seconds{agent_name="main",username="admin",workspace_name="workspace1"} 0.001784
+# HELP coderd_agentstats_currently_reachable_peers The number of peers (e.g. clients) that are currently reachable over the encrypted network.
+# TYPE coderd_agentstats_currently_reachable_peers gauge
+coderd_agentstats_currently_reachable_peers{agent_name="main",connection_type="derp",template_name="docker",username="admin",workspace_name="workspace1"} 0
 # HELP coderd_agentstats_rx_bytes Agent Rx bytes
 # TYPE coderd_agentstats_rx_bytes gauge
 coderd_agentstats_rx_bytes{agent_name="main",username="admin",workspace_name="workspace1"} 7731
