feat: schedule snapshots by CPU compatibility

cocoonset/pods.go

@@ -172,6 +172,10 @@ func toolboxPodName(csName, tbName string) string {
 func newManagedPod(cs *cocoonv1.CocoonSet, podName, role, slotLabel string, scheme *runtime.Scheme) (*corev1.Pod, error) {
 	one := int64(1)
 	pool := cmp.Or(cs.Spec.NodePool, meta.DefaultNodePool)
+	nodeSelector := map[string]string{meta.LabelNodePool: pool}
+	if class := cs.Spec.SnapshotCompatibilityClass; class != "" {
+		nodeSelector[meta.LabelSnapshotCompatibilityClass] = class
+	}
 	pod := &corev1.Pod{
 		ObjectMeta: metav1.ObjectMeta{
 			Name:      podName,
@@ -193,9 +197,7 @@ func newManagedPod(cs *cocoonv1.CocoonSet, podName, role, slotLabel string, sche
 				{Key: corev1.TaintNodeNotReady, Operator: corev1.TolerationOpExists, Effect: corev1.TaintEffectNoExecute},
 				{Key: corev1.TaintNodeUnreachable, Operator: corev1.TolerationOpExists, Effect: corev1.TaintEffectNoExecute},
 			},
-			NodeSelector: map[string]string{
-				meta.LabelNodePool: pool,
-			},
+			NodeSelector: nodeSelector,
 			Containers: []corev1.Container{
 				{
 					Name:  agentContainerName,
@@ -233,7 +235,7 @@ func podSpecMatchesAgent(pod *corev1.Pod, cs *cocoonv1.CocoonSet, slot int32) bo
 	if !equality.Semantic.DeepEqual(pod.Spec.Containers[0].EnvFrom, cs.Spec.Agent.EnvFrom) {
 		return false
 	}
-	if !nodePoolMatches(pod, cs) {
+	if !schedulingMatches(pod, cs) {
 		return false
 	}
 	return true
@@ -247,7 +249,7 @@ func podSpecMatchesToolbox(pod *corev1.Pod, cs *cocoonv1.CocoonSet, tb cocoonv1.
 	if !vmSpecMatches(current, want) || !resourcesMatch(pod, tb.Resources) {
 		return false
 	}
-	if !nodePoolMatches(pod, cs) {
+	if !schedulingMatches(pod, cs) {
 		return false
 	}
 	if tb.Mode == cocoonv1.ToolboxModeStatic {
@@ -298,8 +300,13 @@ func quantityEqual(a, b corev1.ResourceList, name corev1.ResourceName) bool {
 	return qa.Cmp(qb) == 0
 }
 
-func nodePoolMatches(pod *corev1.Pod, cs *cocoonv1.CocoonSet) bool {
-	return meta.PodNodePool(pod) == cmp.Or(cs.Spec.NodePool, meta.DefaultNodePool)
+func schedulingMatches(pod *corev1.Pod, cs *cocoonv1.CocoonSet) bool {
+	if meta.PodNodePool(pod) != cmp.Or(cs.Spec.NodePool, meta.DefaultNodePool) {
+		return false
+	}
+	class := pod.Spec.NodeSelector[meta.LabelSnapshotCompatibilityClass]
+	// NodeSelector is immutable; adopt live pre-feature pods until their next recreate.
+	return class == cs.Spec.SnapshotCompatibilityClass || (class == "" && !podIsTerminal(pod))
 }
 
 // applyStorageRequest propagates the VMOptions.Storage quantity into the

cocoonset/pods_test.go

@@ -176,6 +176,27 @@ func TestNewManagedPodHasOwnerReference(t *testing.T) {
 	}
 }
 
+func TestNewManagedPodSelectsPoolAndSnapshotCompatibilityClass(t *testing.T) {
+	cs := newCocoonSet("demo", func(cs *cocoonv1.CocoonSet) {
+		cs.Spec.NodePool = "purpose-a"
+		cs.Spec.SnapshotCompatibilityClass = "n2-cascade-lake-v1"
+	})
+	pod := mustNewManagedPod(t, cs, "demo-0", meta.RoleMain, "0", testScheme(t))
+	if got := pod.Spec.NodeSelector[meta.LabelNodePool]; got != "purpose-a" {
+		t.Errorf("node pool selector = %q, want purpose-a", got)
+	}
+	if got := pod.Spec.NodeSelector[meta.LabelSnapshotCompatibilityClass]; got != "n2-cascade-lake-v1" {
+		t.Errorf("snapshot compatibility selector = %q, want n2-cascade-lake-v1", got)
+	}
+}
+
+func TestNewManagedPodOmitsEmptySnapshotCompatibilityClass(t *testing.T) {
+	pod := mustNewManagedPod(t, newCocoonSet("demo"), "demo-0", meta.RoleMain, "0", testScheme(t))
+	if _, ok := pod.Spec.NodeSelector[meta.LabelSnapshotCompatibilityClass]; ok {
+		t.Error("legacy CocoonSet must not gain a snapshot compatibility selector")
+	}
+}
+
 func TestNewManagedPodCarriesCocoonToleration(t *testing.T) {
 	cs := newCocoonSet("demo")
 	pod := mustNewManagedPod(t, cs, "demo-0", meta.RoleMain, "0", testScheme(t))
@@ -421,6 +442,32 @@ func TestPodSpecMatchesAgentNodePoolDefaultFallback(t *testing.T) {
 	}
 }
 
+func TestPodSpecMatchesAgentDetectsSnapshotCompatibilityClassDrift(t *testing.T) {
+	cs := newCocoonSet("demo", func(cs *cocoonv1.CocoonSet) {
+		cs.Spec.SnapshotCompatibilityClass = "n2-cascade-lake-v1"
+	})
+	pod := mustBuildAgentPod(t, cs, 0, "", "", testScheme(t))
+	pod.Spec.NodeSelector[meta.LabelSnapshotCompatibilityClass] = "n4-emerald-rapids-v1"
+	if podSpecMatchesAgent(pod, cs, 0) {
+		t.Error("agent pod with the wrong snapshot compatibility selector must not match")
+	}
+}
+
+func TestPodSpecMatchesAgentAdoptsLiveLegacyMissingSnapshotCompatibilityClass(t *testing.T) {
+	cs := newCocoonSet("demo", func(cs *cocoonv1.CocoonSet) {
+		cs.Spec.SnapshotCompatibilityClass = "n2-cascade-lake-v1"
+	})
+	pod := mustBuildAgentPod(t, cs, 0, "", "", testScheme(t))
+	delete(pod.Spec.NodeSelector, meta.LabelSnapshotCompatibilityClass)
+	if !podSpecMatchesAgent(pod, cs, 0) {
+		t.Error("live pre-feature agent pod must be adopted until its next normal recreate")
+	}
+	pod.Status.Phase = corev1.PodFailed
+	if podSpecMatchesAgent(pod, cs, 0) {
+		t.Error("terminal pre-feature agent pod must be recreated with the compatibility selector")
+	}
+}
+
 func TestPodSpecMatchesToolboxDetectsStaticVMIDDrift(t *testing.T) {
 	cs := newCocoonSet("demo")
 	scheme := testScheme(t)
@@ -525,6 +572,18 @@ func TestPodSpecMatchesToolboxDetectsNodePoolDrift(t *testing.T) {
 	}
 }
 
+func TestPodSpecMatchesToolboxAdoptsLegacyMissingSnapshotCompatibilityClass(t *testing.T) {
+	cs := newCocoonSet("demo", func(cs *cocoonv1.CocoonSet) {
+		cs.Spec.SnapshotCompatibilityClass = "n2-cascade-lake-v1"
+	})
+	tb := cocoonv1.ToolboxSpec{Name: "tb", Image: "image", Mode: cocoonv1.ToolboxModeRun}
+	pod := mustBuildToolboxPod(t, cs, tb, testScheme(t))
+	delete(pod.Spec.NodeSelector, meta.LabelSnapshotCompatibilityClass)
+	if !podSpecMatchesToolbox(pod, cs, tb) {
+		t.Error("pre-feature toolbox pod must be adopted until its next normal recreate")
+	}
+}
+
 func TestBuildAgentPodMainPinnedViaHostnameAffinity(t *testing.T) {
 	cs := newCocoonSet("demo", func(cs *cocoonv1.CocoonSet) {
 		cs.Spec.NodeName = "node-b"

cocoonset/slotrelease_test.go

@@ -164,6 +164,7 @@ func TestSuspendReleaseFreesSeatWhenFlagPatchFails(t *testing.T) {
 func TestWakeRecreatesWithRestoreAndPreferredAffinity(t *testing.T) {
 	cs := relCocoonSet(func(cs *cocoonv1.CocoonSet) {
 		cs.Spec.Suspend = false
+		cs.Spec.SnapshotCompatibilityClass = "n2-cascade-lake-v1"
 		cs.Status.Phase = cocoonv1.CocoonSetPhaseSuspended
 		cs.Annotations = map[string]string{meta.AnnotationHibernatedOnNode: "node-a"}
 	})
@@ -185,6 +186,9 @@ func TestWakeRecreatesWithRestoreAndPreferredAffinity(t *testing.T) {
 	if got.Spec.NodeName != "" {
 		t.Errorf("wake must not hard-pin a node, got %q", got.Spec.NodeName)
 	}
+	if got.Spec.NodeSelector[meta.LabelSnapshotCompatibilityClass] != "n2-cascade-lake-v1" {
+		t.Errorf("wake must preserve snapshot compatibility selector, got %v", got.Spec.NodeSelector)
+	}
 	na := got.Spec.Affinity
 	if na == nil || na.NodeAffinity == nil || len(na.NodeAffinity.PreferredDuringSchedulingIgnoredDuringExecution) != 1 ||
 		na.NodeAffinity.PreferredDuringSchedulingIgnoredDuringExecution[0].Preference.MatchExpressions[0].Values[0] != "node-a" {

config/crd/bases/cocoonset.cocoonstack.io_cocoonsets.yaml

@@ -267,6 +267,14 @@ spec:
               nodePool:
                 default: default
                 type: string
+              snapshotCompatibilityClass:
+                description: |-
+                  SnapshotCompatibilityClass selects nodes that expose a guest-visible CPU
+                  ABI compatible with this set's memory snapshots. It is independent from
+                  NodePool so multiple workload pools can share one snapshot class.
+                maxLength: 63
+                pattern: ^(([A-Za-z0-9][-A-Za-z0-9_.]*)?[A-Za-z0-9])?$
+                type: string
               snapshotPolicy:
                 default: always
                 description: SnapshotPolicy defines when VM snapshots are taken.
@@ -436,6 +444,9 @@ spec:
               rule: '!has(self.hibernatePolicy) || self.hibernatePolicy != ''release''
                 || ((!has(self.agent.replicas) || self.agent.replicas == 0) && (!has(self.toolboxes)
                 || size(self.toolboxes) == 0))'
+            - message: snapshotCompatibilityClass is immutable once set
+              rule: '!has(oldSelf.snapshotCompatibilityClass) || (has(self.snapshotCompatibilityClass)
+                && self.snapshotCompatibilityClass == oldSelf.snapshotCompatibilityClass)'
           status:
             description: CocoonSetStatus represents the observed state of a CocoonSet.
             properties:

docs/cocoonset.md

@@ -16,6 +16,8 @@
 
 Pods are constructed via `meta.FromAgentSpec` / `meta.FromToolboxSpec` factory helpers so the operator never touches the annotation map directly. These factories propagate the full `VMOptions` surface (OS, Backend, ConnType, Network, ForcePull, NoDirectIO, ProbePort, Storage, Resources) into the pod annotations that vk-cocoon consumes. The `For` watch uses `predicate.GenerationChangedPredicate` so reconciles only fire when the spec actually changes — status-only patches the operator makes itself never loop back. The `Owns` side filters pod events to creation, deletion, and meaningful transitions (phase change, readiness flip, label/annotation mutation) via a `podRelevantChange` predicate so pure VK status churn does not trigger reconcile storms.
 
+`spec.nodePool` and `spec.snapshotCompatibilityClass` are independent hard placement dimensions. Every managed pod selects `cocoonstack.io/pool=<nodePool>`; a non-empty snapshot class also selects `cocoonstack.io/snapshot-cpu-class=<class>`. The class identifies a certified guest-visible CPU ABI, not merely `amd64`, and remains in force across release-policy wake, rebuild, and migration. Hostname affinity stays a locality preference or explicit migration target and cannot relax either selector. Pre-feature pods whose immutable `nodeSelector` lacks the class are adopted rather than drift-deleted; their next normal recreate or released-seat wake converges them onto the hard selector.
+
 See [Observability](observability.md) for the Event reasons and metrics
 this loop emits, and [Configuration](configuration.md) for the operator's
 environment variables.

go.mod

@@ -3,7 +3,7 @@ module github.com/cocoonstack/cocoon-operator
 go 1.26.5
 
 require (
-	github.com/cocoonstack/cocoon-common v0.2.9-0.20260731042413-9ca4f1c8fc0f
+	github.com/cocoonstack/cocoon-common v0.2.10-0.20260811071104-34821ee9f5ef
 	github.com/go-logr/logr v1.4.3
 	github.com/google/go-containerregistry v0.21.7
 	github.com/projecteru2/core v0.0.0-20241016125006-ff909eefe04c

go.sum

@@ -29,8 +29,8 @@ github.com/cockroachdb/logtags v0.0.0-20230118201751-21c54148d20b h1:r6VH0faHjZe
 github.com/cockroachdb/logtags v0.0.0-20230118201751-21c54148d20b/go.mod h1:Vz9DsVWQQhf3vs21MhPMZpMGSht7O/2vFW2xusFUVOs=
 github.com/cockroachdb/redact v1.1.3 h1:AKZds10rFSIj7qADf0g46UixK8NNLwWTNdCIGS5wfSQ=
 github.com/cockroachdb/redact v1.1.3/go.mod h1:BVNblN9mBWFyMyqK1k3AAiSxhvhfK2oOZZ2lK+dpvRg=
-github.com/cocoonstack/cocoon-common v0.2.9-0.20260731042413-9ca4f1c8fc0f h1:mcHAOv4VUN1cc3CSAeoWyKh/iwPQ4WithgB2sKAVi0Y=
-github.com/cocoonstack/cocoon-common v0.2.9-0.20260731042413-9ca4f1c8fc0f/go.mod h1:VSfgYiWxoHRnWybzQNaxmK4kVoLT9ffiMKll5/i92CM=
+github.com/cocoonstack/cocoon-common v0.2.10-0.20260811071104-34821ee9f5ef h1:715n/BrVpsRBzPvj+GQyU2xrVV7LiLOf1qBhRuRDN6U=
+github.com/cocoonstack/cocoon-common v0.2.10-0.20260811071104-34821ee9f5ef/go.mod h1:VSfgYiWxoHRnWybzQNaxmK4kVoLT9ffiMKll5/i92CM=
 github.com/codegangsta/inject v0.0.0-20150114235600-33e0aa1cb7c0/go.mod h1:4Zcjuz89kmFXt9morQgcfYZAYZ5n8WHjt81YYWIwtTM=
 github.com/coreos/etcd v3.3.10+incompatible/go.mod h1:uF7uidLiAD3TWHmW31ZFd/JWoc32PjwdhPthX9715RE=
 github.com/coreos/go-etcd v2.0.0+incompatible/go.mod h1:Jez6KQU2B/sWsbdaef3ED8NzMklzPG4d5KIOhIy30Tk=