Merge pull request #6003 from pete-woods/add-job-status-metrics

nomad: add job status metrics

nomad/leader.go

@@ -254,6 +254,9 @@ func (s *Server) establishLeadership(stopCh chan struct{}) error {
 	// Periodically publish job summary metrics
 	go s.publishJobSummaryMetrics(stopCh)
 
+	// Periodically publish job status metrics
+	go s.publishJobStatusMetrics(stopCh)
+
 	// Setup the heartbeat timers. This is done both when starting up or when
 	// a leader fail over happens. Since the timers are maintained by the leader
 	// node, effectively this means all the timers are renewed at the time of failover.
@@ -700,6 +703,62 @@ func (s *Server) iterateJobSummaryMetrics(summary *structs.JobSummary) {
 	}
 }
 
+// publishJobStatusMetrics publishes the job statuses as metrics
+func (s *Server) publishJobStatusMetrics(stopCh chan struct{}) {
+	timer := time.NewTimer(0)
+	defer timer.Stop()
+
+	for {
+		select {
+		case <-stopCh:
+			return
+		case <-timer.C:
+			timer.Reset(s.config.StatsCollectionInterval)
+			state, err := s.State().Snapshot()
+			if err != nil {
+				s.logger.Error("failed to get state", "error", err)
+				continue
+			}
+			ws := memdb.NewWatchSet()
+			iter, err := state.Jobs(ws)
+			if err != nil {
+				s.logger.Error("failed to get job statuses", "error", err)
+				continue
+			}
+
+			s.iterateJobStatusMetrics(&iter)
+		}
+	}
+}
+
+func (s *Server) iterateJobStatusMetrics(jobs *memdb.ResultIterator) {
+	var pending int64 // Sum of all jobs in 'pending' state
+	var running int64 // Sum of all jobs in 'running' state
+	var dead int64    // Sum of all jobs in 'dead' state
+
+	for {
+		raw := (*jobs).Next()
+		if raw == nil {
+			break
+		}
+
+		job := raw.(*structs.Job)
+
+		switch job.Status {
+		case structs.JobStatusPending:
+			pending++
+		case structs.JobStatusRunning:
+			running++
+		case structs.JobStatusDead:
+			dead++
+		}
+	}
+
+	metrics.SetGauge([]string{"nomad", "job_status", "pending"}, float32(pending))
+	metrics.SetGauge([]string{"nomad", "job_status", "running"}, float32(running))
+	metrics.SetGauge([]string{"nomad", "job_status", "dead"}, float32(dead))
+}
+
 // revokeLeadership is invoked once we step down as leader.
 // This is used to cleanup any state that may be specific to a leader.
 func (s *Server) revokeLeadership() error {

website/source/docs/telemetry/metrics.html.md

@@ -705,9 +705,9 @@ detailed above) but any new metrics will only be available in the new format.
   </tr>
 </table>
 
-## Job Metrics
+## Job Summary Metrics
 
-Job metrics are emitted by the Nomad leader server.
+Job summary metrics are emitted by the Nomad leader server.
 
 <table class="table table-bordered table-striped">
   <tr>
@@ -761,6 +761,37 @@ Job metrics are emitted by the Nomad leader server.
   </tr>
 </table>
 
+## Job Status Metrics
+
+Job status metrics are emitted by the Nomad leader server.
+
+<table class="table table-bordered table-striped">
+  <tr>
+    <th>Metric</th>
+    <th>Description</th>
+    <th>Unit</th>
+    <th>Type</th>
+  </tr>
+  <tr>
+    <td>`nomad.job_status.pending`</td>
+    <td>Number jobs pending</td>
+    <td>Integer</td>
+    <td>Gauge</td>
+  </tr>
+  <tr>
+    <td>`nomad.job_status.running`</td>
+    <td>Number jobs running</td>
+    <td>Integer</td>
+    <td>Gauge</td>
+  </tr>
+  <tr>
+    <td>`nomad.job_status.dead`</td>
+    <td>Number of dead jobs</td>
+    <td>Integer</td>
+    <td>Gauge</td>
+  </tr>
+</table>
+
 ## Metric Types
 
 <table class="table table-bordered table-striped">