feat: collect container requests/limits and pod QoS class

cmd/agent/main.go

@@ -71,6 +71,7 @@ func run(ctx context.Context, logger *slog.Logger, clientset kubernetes.Interfac
 			"pod", p.Name,
 			"node", p.Node,
 			"phase", p.Phase,
+			"qos", p.QOSClass,
 			"workload_kind", p.Workload.Kind,
 			"workload_name", p.Workload.Name,
 			"containers", p.Containers,

internal/collector/pods.go

@@ -13,6 +13,7 @@ import (
 	appslisters "k8s.io/client-go/listers/apps/v1"
 	batchlisters "k8s.io/client-go/listers/batch/v1"
 	"k8s.io/client-go/tools/cache"
+	"k8s.io/utils/ptr"
 )
 
 // WorkloadRef identifies the controller that ultimately manages a pod:
@@ -24,12 +25,25 @@ type WorkloadRef struct {
 	Name string `json:"name"`
 }
 
-// Container is the collected view of a container: name and image only.
-// Env, args, and command are deliberately never read (filter early).
+// Resources is the declared resource envelope of one container, normalized
+// to resource units: CPU in millicores, memory in bytes. A nil field means
+// the corresponding request or limit is not set — a meaningful fact in
+// itself, distinct from zero.
+type Resources struct {
+	CPURequestMilli    *int64 `json:"cpu_request_milli,omitempty"`
+	CPULimitMilli      *int64 `json:"cpu_limit_milli,omitempty"`
+	MemoryRequestBytes *int64 `json:"memory_request_bytes,omitempty"`
+	MemoryLimitBytes   *int64 `json:"memory_limit_bytes,omitempty"`
+}
+
+// Container is the collected view of a container: name, image, and declared
+// resources only. Env, args, and command are deliberately never read
+// (filter early).
 type Container struct {
-	Name  string `json:"name"`
-	Image string `json:"image"`
-	Init  bool   `json:"init,omitempty"`
+	Name      string    `json:"name"`
+	Image     string    `json:"image"`
+	Init      bool      `json:"init,omitempty"`
+	Resources Resources `json:"resources"`
 }
 
 // PodInfo is the collected view of one pod.
@@ -38,6 +52,7 @@ type PodInfo struct {
 	Name       string
 	Node       string
 	Phase      string
+	QOSClass   string
 	Workload   WorkloadRef
 	Containers []Container
 }
@@ -111,17 +126,39 @@ func (w *PodWatcher) describe(pod *corev1.Pod) PodInfo {
 		Name:      pod.Name,
 		Node:      pod.Spec.NodeName,
 		Phase:     string(pod.Status.Phase),
+		QOSClass:  string(pod.Status.QOSClass),
 		Workload:  w.resolveWorkload(pod),
 	}
 	for _, c := range pod.Spec.InitContainers {
-		info.Containers = append(info.Containers, Container{Name: c.Name, Image: c.Image, Init: true})
+		info.Containers = append(info.Containers,
+			Container{Name: c.Name, Image: c.Image, Init: true, Resources: resourcesOf(&c)})
 	}
 	for _, c := range pod.Spec.Containers {
-		info.Containers = append(info.Containers, Container{Name: c.Name, Image: c.Image})
+		info.Containers = append(info.Containers,
+			Container{Name: c.Name, Image: c.Image, Resources: resourcesOf(&c)})
 	}
 	return info
 }
 
+// resourcesOf normalizes a container's declared requests and limits:
+// CPU quantities to millicores, memory to bytes. Absent entries stay nil.
+func resourcesOf(c *corev1.Container) Resources {
+	var r Resources
+	if q, ok := c.Resources.Requests[corev1.ResourceCPU]; ok {
+		r.CPURequestMilli = ptr.To(q.MilliValue())
+	}
+	if q, ok := c.Resources.Limits[corev1.ResourceCPU]; ok {
+		r.CPULimitMilli = ptr.To(q.MilliValue())
+	}
+	if q, ok := c.Resources.Requests[corev1.ResourceMemory]; ok {
+		r.MemoryRequestBytes = ptr.To(q.Value())
+	}
+	if q, ok := c.Resources.Limits[corev1.ResourceMemory]; ok {
+		r.MemoryLimitBytes = ptr.To(q.Value())
+	}
+	return r
+}
+
 // resolveWorkload walks the owner chain one controller hop at a time:
 // Pod → ReplicaSet → Deployment (or Argo Rollout), Pod → Job → CronJob.
 // Direct controllers (StatefulSet, DaemonSet, custom CRDs) are reported

internal/collector/pods_test.go

@@ -2,13 +2,15 @@ package collector
 
 import (
 	"context"
+	"fmt"
 	"sync"
 	"testing"
 	"time"
 
 	appsv1 "k8s.io/api/apps/v1"
 	batchv1 "k8s.io/api/batch/v1"
 	corev1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/resource"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
 	"k8s.io/apimachinery/pkg/types"
@@ -29,14 +31,25 @@ func pod(name string, owner *metav1.OwnerReference) *corev1.Pod {
 		Spec: corev1.PodSpec{
 			NodeName: "node-1",
 			InitContainers: []corev1.Container{
-				{Name: "init-db", Image: "example.com/migrate:v3"},
+				{Name: "init-db", Image: "example.com/migrate:v3", Resources: corev1.ResourceRequirements{
+					Requests: corev1.ResourceList{corev1.ResourceCPU: resource.MustParse("100m")},
+				}},
 			},
 			Containers: []corev1.Container{
-				{Name: "app", Image: "example.com/app:1.2.3"},
+				{Name: "app", Image: "example.com/app:1.2.3", Resources: corev1.ResourceRequirements{
+					Requests: corev1.ResourceList{
+						corev1.ResourceCPU:    resource.MustParse("500m"),
+						corev1.ResourceMemory: resource.MustParse("256Mi"),
+					},
+					Limits: corev1.ResourceList{
+						corev1.ResourceCPU:    resource.MustParse("2"),
+						corev1.ResourceMemory: resource.MustParse("1Gi"),
+					},
+				}},
 				{Name: "sidecar", Image: "example.com/proxy:latest"},
 			},
 		},
-		Status: corev1.PodStatus{Phase: corev1.PodRunning},
+		Status: corev1.PodStatus{Phase: corev1.PodRunning, QOSClass: corev1.PodQOSBurstable},
 	}
 	if owner != nil {
 		p.OwnerReferences = []metav1.OwnerReference{*owner}
@@ -137,21 +150,58 @@ func TestCollectsContainersAndImages(t *testing.T) {
 
 	info := seen["shop/checkout-1"]
 	want := []Container{
-		{Name: "init-db", Image: "example.com/migrate:v3", Init: true},
-		{Name: "app", Image: "example.com/app:1.2.3"},
+		{Name: "init-db", Image: "example.com/migrate:v3", Init: true,
+			Resources: Resources{CPURequestMilli: ptr.To(int64(100))}},
+		{Name: "app", Image: "example.com/app:1.2.3",
+			Resources: Resources{
+				CPURequestMilli:    ptr.To(int64(500)),
+				CPULimitMilli:      ptr.To(int64(2000)),
+				MemoryRequestBytes: ptr.To(int64(256 << 20)),
+				MemoryLimitBytes:   ptr.To(int64(1 << 30)),
+			}},
 		{Name: "sidecar", Image: "example.com/proxy:latest"},
 	}
 	if len(info.Containers) != len(want) {
 		t.Fatalf("containers = %+v, want %+v", info.Containers, want)
 	}
 	for i := range want {
-		if info.Containers[i] != want[i] {
-			t.Errorf("container %d = %+v, want %+v", i, info.Containers[i], want[i])
+		got := info.Containers[i]
+		if got.Name != want[i].Name || got.Image != want[i].Image || got.Init != want[i].Init {
+			t.Errorf("container %d = %+v, want %+v", i, got, want[i])
 		}
+		assertResources(t, got.Name, got.Resources, want[i].Resources)
 	}
 	if info.Node != "node-1" || info.Phase != string(corev1.PodRunning) {
 		t.Errorf("node/phase = %q/%q, want node-1/Running", info.Node, info.Phase)
 	}
+	if info.QOSClass != string(corev1.PodQOSBurstable) {
+		t.Errorf("qos = %q, want Burstable", info.QOSClass)
+	}
+}
+
+// assertResources compares two Resources field by field; a nil on either
+// side must be nil on the other (absent means "not set", not zero).
+func assertResources(t *testing.T, container string, got, want Resources) {
+	t.Helper()
+	check := func(field string, g, w *int64) {
+		switch {
+		case (g == nil) != (w == nil):
+			t.Errorf("%s: %s = %v, want %v", container, field, format(g), format(w))
+		case g != nil && *g != *w:
+			t.Errorf("%s: %s = %d, want %d", container, field, *g, *w)
+		}
+	}
+	check("cpu_request_milli", got.CPURequestMilli, want.CPURequestMilli)
+	check("cpu_limit_milli", got.CPULimitMilli, want.CPULimitMilli)
+	check("memory_request_bytes", got.MemoryRequestBytes, want.MemoryRequestBytes)
+	check("memory_limit_bytes", got.MemoryLimitBytes, want.MemoryLimitBytes)
+}
+
+func format(v *int64) string {
+	if v == nil {
+		return "unset"
+	}
+	return fmt.Sprintf("%d", *v)
 }
 
 func TestReportsPodsCreatedAfterStart(t *testing.T) {

test/e2e/pods_test.go

@@ -18,6 +18,7 @@ import (
 	appsv1 "k8s.io/api/apps/v1"
 	batchv1 "k8s.io/api/batch/v1"
 	corev1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/resource"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/client-go/kubernetes"
 	"k8s.io/client-go/tools/clientcmd"
@@ -77,7 +78,19 @@ func TestPodWatcherAgainstRealCluster(t *testing.T) {
 			Template: corev1.PodTemplateSpec{
 				ObjectMeta: metav1.ObjectMeta{Labels: map[string]string{"app": "web"}},
 				Spec: corev1.PodSpec{
-					Containers: []corev1.Container{{Name: "app", Image: pauseImage}},
+					Containers: []corev1.Container{{
+						Name: "app", Image: pauseImage,
+						Resources: corev1.ResourceRequirements{
+							Requests: corev1.ResourceList{
+								corev1.ResourceCPU:    resource.MustParse("50m"),
+								corev1.ResourceMemory: resource.MustParse("32Mi"),
+							},
+							Limits: corev1.ResourceList{
+								corev1.ResourceCPU:    resource.MustParse("100m"),
+								corev1.ResourceMemory: resource.MustParse("64Mi"),
+							},
+						},
+					}},
 				},
 			},
 		},
@@ -154,6 +167,24 @@ func TestPodWatcherAgainstRealCluster(t *testing.T) {
 	if len(webPod.Containers) != 1 || webPod.Containers[0].Image != pauseImage {
 		t.Errorf("deployment pod containers = %+v, want one %s", webPod.Containers, pauseImage)
 	}
+	res := webPod.Containers[0].Resources
+	wantRes := map[string]struct {
+		got  *int64
+		want int64
+	}{
+		"cpu_request_milli":    {res.CPURequestMilli, 50},
+		"cpu_limit_milli":      {res.CPULimitMilli, 100},
+		"memory_request_bytes": {res.MemoryRequestBytes, 32 << 20},
+		"memory_limit_bytes":   {res.MemoryLimitBytes, 64 << 20},
+	}
+	for field, v := range wantRes {
+		if v.got == nil || *v.got != v.want {
+			t.Errorf("deployment pod %s = %v, want %d", field, v.got, v.want)
+		}
+	}
+	if webPod.QOSClass != string(corev1.PodQOSBurstable) {
+		t.Errorf("deployment pod qos = %q, want Burstable", webPod.QOSClass)
+	}
 
 	waitForWorkload("CronJob/tick")