Merge pull request #10 from smartnews/spot-metrics

chore: add hostname to spot int metrics
smartnews · May 14, 2024 · 1310e4d · 1310e4d
2 parents 0026970 + a7afc03
commit 1310e4d
Show file tree

Hide file tree

Showing 2 changed files with 4 additions and 2 deletions.
diff --git a/pkg/controllers/interruption/controller.go b/pkg/controllers/interruption/controller.go
@@ -198,7 +198,7 @@ func (c *Controller) handleNodeClaim(ctx context.Context, msg messages.Message,
 		if zone != "" && instanceType != "" {
 			c.unavailableOfferingsCache.MarkUnavailable(ctx, string(msg.Kind()), instanceType, zone, v1beta1.CapacityTypeSpot)
 		}
-		spotTotal.WithLabelValues(instanceType, zone).Inc()
+		spotTotal.WithLabelValues(instanceType, zone, nodeClaim.Status.NodeName, nodeClaim.Labels["karpenter.sh/nodepool"]).Inc()
 		// try to create a new nodeclaim immediately but ignore error if it fails
 		if err := c.createNodeClaim(ctx, nodeClaim); err != nil {
 			logging.FromContext(ctx).Errorf("[interruption handling]failed to create a new nodeclaim, %v", err)

diff --git a/pkg/controllers/interruption/metrics.go b/pkg/controllers/interruption/metrics.go
@@ -28,6 +28,8 @@ const (
 	terminationReasonLabel = "interruption"
 	instanceTypeLabel      = "instance_type"
 	zoneLabel              = "zone"
+	hostLabel              = "node_name"
+	poolLabel              = "node_pool"
 )
 
 var (
@@ -73,7 +75,7 @@ var (
 			Name:      "spot_total",
 			Help:      "Number of the spot interruption. Labeled by AZ, instance type",
 		},
-		[]string{instanceTypeLabel, zoneLabel},
+		[]string{instanceTypeLabel, zoneLabel, hostLabel, poolLabel},
 	)
 )