feat: add workspace build timing metrics (coder#15771)

kevinh-canva · web-flow · commit c5287910f957 · 2024-12-11T05:36:48.000Z
This PR introduces a new prometheus metrics for
`workspace_build_timing_seconds`, which specifically reports workspace
build times. To reduce cardinality, this metrics excludes
`workspace_name` and `workspace_owner` that are present on the
`workspace_builds_total` metrics.
diff --git a/provisionerd/provisionerd.go b/provisionerd/provisionerd.go
@@ -178,6 +178,22 @@ func NewMetrics(reg prometheus.Registerer) Metrics {
 				Name:      "workspace_builds_total",
 				Help:      "The number of workspaces started, updated, or deleted.",
 			}, []string{"workspace_owner", "workspace_name", "template_name", "template_version", "workspace_transition", "status"}),
+			WorkspaceBuildTimings: auto.NewHistogramVec(prometheus.HistogramOpts{
+				Namespace: "coderd",
+				Subsystem: "provisionerd",
+				Name:      "workspace_build_timings_seconds",
+				Help:      "The time taken for a workspace to build.",
+				Buckets: []float64{
+					1, // 1s
+					10,
+					30,
+					60, // 1min
+					60 * 5,
+					60 * 10,
+					60 * 30, // 30min
+					60 * 60, // 1hr
+				},
+			}, []string{"template_name", "template_version", "workspace_transition", "status"}),
 		},
 	}
 }
diff --git a/provisionerd/runner/runner.go b/provisionerd/runner/runner.go
@@ -86,7 +86,8 @@ type Metrics struct {
 	// JobTimings also counts the total amount of jobs.
 	JobTimings *prometheus.HistogramVec
 	// WorkspaceBuilds counts workspace build successes and failures.
-	WorkspaceBuilds *prometheus.CounterVec
+	WorkspaceBuilds       *prometheus.CounterVec
+	WorkspaceBuildTimings *prometheus.HistogramVec
 }
 
 type JobUpdater interface {
@@ -189,6 +190,12 @@ func (r *Runner) Run() {
 				build.Metadata.WorkspaceTransition.String(),
 				status,
 			).Inc()
+			r.metrics.WorkspaceBuildTimings.WithLabelValues(
+				build.Metadata.TemplateName,
+				build.Metadata.TemplateVersion,
+				build.Metadata.WorkspaceTransition.String(),
+				status,
+			).Observe(time.Since(start).Seconds())
 		}
 		r.metrics.JobTimings.WithLabelValues(r.job.Provisioner, status).Observe(time.Since(start).Seconds())
 	}()
