WIP: Implementation 5800 kep exclude resources

apis/kueue/v1beta1/clusterqueue_types.go

@@ -142,6 +142,24 @@ type ClusterQueueSpec struct {
 	// admissionScope indicates whether ClusterQueue uses the Admission Fair Sharing
 	// +optional
 	AdmissionScope *AdmissionScope `json:"admissionScope,omitempty"`
+
+	// excludeResourcePrefixes defines which resources should be ignored by
+	// Kueue for quota management in this ClusterQueue. Resources matching any
+	// of the prefixes will be excluded from quota calculations.
+	// When specified, this list is combined (union) with the global
+	// excludeResourcePrefixes from the Kueue Configuration.
+	// The prefix matching follows the same semantics as the global configuration:
+	// - An exact match of the resource name
+	// - A prefix match followed by a slash (e.g., "example.com" matches "example.com/gpu")
+	//
+	// Example: ["ephemeral-storage", "hugepages-", "example.com"]
+	//
+	// +listType=set
+	// +kubebuilder:validation:MaxItems=64
+	// +kubebuilder:validation:items:MaxLength=256
+	// +kubebuilder:validation:items:MinLength=1
+	// +optional
+	ExcludeResourcePrefixes []string `json:"excludeResourcePrefixes,omitempty"`
 }
 
 // AdmissionChecksStrategy defines a strategy for a AdmissionCheck.

apis/kueue/v1beta2/clusterqueue_types.go

@@ -142,6 +142,24 @@ type ClusterQueueSpec struct {
 	// admissionScope indicates whether ClusterQueue uses the Admission Fair Sharing
 	// +optional
 	AdmissionScope *AdmissionScope `json:"admissionScope,omitempty"`
+
+	// excludeResourcePrefixes defines which resources should be ignored by
+	// Kueue for quota management in this ClusterQueue. Resources matching any
+	// of the prefixes will be excluded from quota calculations.
+	// When specified, this list is combined (union) with the global
+	// excludeResourcePrefixes from the Kueue Configuration.
+	// The prefix matching follows the same semantics as the global configuration:
+	// - An exact match of the resource name
+	// - A prefix match followed by a slash (e.g., "example.com" matches "example.com/gpu")
+	//
+	// Example: ["ephemeral-storage", "hugepages-", "example.com"]
+	//
+	// +listType=set
+	// +kubebuilder:validation:MaxItems=64
+	// +kubebuilder:validation:items:MaxLength=256
+	// +kubebuilder:validation:items:MinLength=1
+	// +optional
+	ExcludeResourcePrefixes []string `json:"excludeResourcePrefixes,omitempty"`
 }
 
 // AdmissionChecksStrategy defines a strategy for a AdmissionCheck.

config/components/crd/bases/kueue.x-k8s.io_clusterqueues.yaml

@@ -131,6 +131,25 @@ spec:
                 maxLength: 253
                 pattern: ^[a-z0-9]([-a-z0-9]*[a-z0-9])?(\.[a-z0-9]([-a-z0-9]*[a-z0-9])?)*$
                 type: string
+              excludeResourcePrefixes:
+                description: |-
+                  excludeResourcePrefixes defines which resources should be ignored by
+                  Kueue for quota management in this ClusterQueue. Resources matching any
+                  of the prefixes will be excluded from quota calculations.
+                  When specified, this list is combined (union) with the global
+                  excludeResourcePrefixes from the Kueue Configuration.
+                  The prefix matching follows the same semantics as the global configuration:
+                  - An exact match of the resource name
+                  - A prefix match followed by a slash (e.g., "example.com" matches "example.com/gpu")
+
+                  Example: ["ephemeral-storage", "hugepages-", "example.com"]
+                items:
+                  maxLength: 256
+                  minLength: 1
+                  type: string
+                maxItems: 64
+                type: array
+                x-kubernetes-list-type: set
               fairSharing:
                 description: |-
                   fairSharing defines the properties of the ClusterQueue when
@@ -894,6 +913,25 @@ spec:
                 maxLength: 253
                 pattern: ^[a-z0-9]([-a-z0-9]*[a-z0-9])?(\.[a-z0-9]([-a-z0-9]*[a-z0-9])?)*$
                 type: string
+              excludeResourcePrefixes:
+                description: |-
+                  excludeResourcePrefixes defines which resources should be ignored by
+                  Kueue for quota management in this ClusterQueue. Resources matching any
+                  of the prefixes will be excluded from quota calculations.
+                  When specified, this list is combined (union) with the global
+                  excludeResourcePrefixes from the Kueue Configuration.
+                  The prefix matching follows the same semantics as the global configuration:
+                  - An exact match of the resource name
+                  - A prefix match followed by a slash (e.g., "example.com" matches "example.com/gpu")
+
+                  Example: ["ephemeral-storage", "hugepages-", "example.com"]
+                items:
+                  maxLength: 256
+                  minLength: 1
+                  type: string
+                maxItems: 64
+                type: array
+                x-kubernetes-list-type: set
               fairSharing:
                 description: |-
                   fairSharing defines the properties of the ClusterQueue when

pkg/cache/queue/cluster_queue.go

@@ -114,12 +114,24 @@ type ClusterQueue struct {
 	localQueuesInClusterQueue map[utilqueue.LocalQueueReference]bool
 
 	sw *stickyWorkload
+
+	// excludeResourcePrefixes contains the resource prefixes to exclude from quota management
+	// at the ClusterQueue level. These are merged with global exclusions.
+	excludeResourcePrefixes []string
 }
 
 func (c *ClusterQueue) GetName() kueue.ClusterQueueReference {
 	return c.name
 }
 
+// GetExcludeResourcePrefixes returns the ClusterQueue-specific resource exclusion prefixes.
+// These should be merged with global exclusions when creating workload Info.
+func (c *ClusterQueue) GetExcludeResourcePrefixes() []string {
+	c.rwm.RLock()
+	defer c.rwm.RUnlock()
+	return slices.Clone(c.excludeResourcePrefixes)
+}
+
 func workloadKey(i *workload.Info) workload.Reference {
 	return workload.Key(i.Obj)
 }
@@ -162,6 +174,7 @@ func (c *ClusterQueue) Update(apiCQ *kueue.ClusterQueue) error {
 	}
 	c.namespaceSelector = nsSelector
 	c.active = apimeta.IsStatusConditionTrue(apiCQ.Status.Conditions, kueue.ClusterQueueActive)
+	c.excludeResourcePrefixes = slices.Clone(apiCQ.Spec.ExcludeResourcePrefixes)
 	return nil
 }
 

pkg/cache/queue/cluster_queue_test.go

@@ -1138,3 +1138,111 @@ func TestFsAdmission(t *testing.T) {
 		})
 	}
 }
+
+func TestClusterQueueGetExcludeResourcePrefixes(t *testing.T) {
+	ctx, _ := utiltesting.ContextWithLog(t)
+
+	cases := map[string]struct {
+		cqExclusions []string
+		want         []string
+	}{
+		"no exclusions": {
+			cqExclusions: []string{},
+			want:         []string{},
+		},
+		"single exclusion": {
+			cqExclusions: []string{"example.com/"},
+			want:         []string{"example.com/"},
+		},
+		"multiple exclusions": {
+			cqExclusions: []string{"example.com/", "foo.io/", "bar.io/"},
+			want:         []string{"example.com/", "foo.io/", "bar.io/"},
+		},
+	}
+
+	for name, tc := range cases {
+		t.Run(name, func(t *testing.T) {
+			cq := newClusterQueueImpl(ctx, nil, defaultOrdering, testingclock.NewFakeClock(time.Now()), nil, false, nil)
+			apiCQ := utiltestingapi.MakeClusterQueue("test-cq").
+				ExcludeResourcePrefixes(tc.cqExclusions).
+				Obj()
+
+			err := cq.Update(apiCQ)
+			if err != nil {
+				t.Fatalf("Failed to update ClusterQueue: %v", err)
+			}
+
+			got := cq.GetExcludeResourcePrefixes()
+			if diff := cmp.Diff(tc.want, got); diff != "" {
+				t.Errorf("Unexpected exclude resource prefixes (-want,+got):\n%s", diff)
+			}
+
+			// Verify that the returned slice is a copy (not the original)
+			if len(got) > 0 {
+				got[0] = "modified/"
+				unmodified := cq.GetExcludeResourcePrefixes()
+				if diff := cmp.Diff(tc.want, unmodified); diff != "" {
+					t.Errorf("Modifying returned slice affected internal state (-want,+got):\n%s", diff)
+				}
+			}
+		})
+	}
+}
+
+func TestClusterQueueUpdateExclusions(t *testing.T) {
+	ctx, _ := utiltesting.ContextWithLog(t)
+
+	cases := map[string]struct {
+		initialExclusions []string
+		updatedExclusions []string
+	}{
+		"empty to non-empty": {
+			initialExclusions: []string{},
+			updatedExclusions: []string{"example.com/"},
+		},
+		"non-empty to empty": {
+			initialExclusions: []string{"example.com/"},
+			updatedExclusions: []string{},
+		},
+		"update existing": {
+			initialExclusions: []string{"example.com/"},
+			updatedExclusions: []string{"foo.io/", "bar.io/"},
+		},
+		"no change": {
+			initialExclusions: []string{"example.com/"},
+			updatedExclusions: []string{"example.com/"},
+		},
+	}
+
+	for name, tc := range cases {
+		t.Run(name, func(t *testing.T) {
+			cq := newClusterQueueImpl(ctx, nil, defaultOrdering, testingclock.NewFakeClock(time.Now()), nil, false, nil)
+
+			// Set initial exclusions
+			apiCQ := utiltestingapi.MakeClusterQueue("test-cq").
+				ExcludeResourcePrefixes(tc.initialExclusions).
+				Obj()
+			err := cq.Update(apiCQ)
+			if err != nil {
+				t.Fatalf("Failed initial update: %v", err)
+			}
+
+			got := cq.GetExcludeResourcePrefixes()
+			if diff := cmp.Diff(tc.initialExclusions, got); diff != "" {
+				t.Errorf("Initial exclusions incorrect (-want,+got):\n%s", diff)
+			}
+
+			// Update exclusions
+			apiCQ.Spec.ExcludeResourcePrefixes = tc.updatedExclusions
+			err = cq.Update(apiCQ)
+			if err != nil {
+				t.Fatalf("Failed to update ClusterQueue: %v", err)
+			}
+
+			got = cq.GetExcludeResourcePrefixes()
+			if diff := cmp.Diff(tc.updatedExclusions, got); diff != "" {
+				t.Errorf("Updated exclusions incorrect (-want,+got):\n%s", diff)
+			}
+		})
+	}
+}