docs: 保留 DBaaS 已批准设计合同

.agents/skills/homelab-knowledge/SKILL.md

@@ -1,76 +0,0 @@
----
-name: homelab-knowledge
-description: Query and maintain the shared homelab-wiki when working on homelab services, infrastructure, architecture, operations, or current service status. Use it to gather existing context before work and to keep durable knowledge synchronized after relevant changes; do not use it for unrelated software work or as a substitute for commit and PR history.
----
-
-# Homelab Knowledge
-
-Use `homelab-wiki` as the shared long-lived knowledge base for people and agents. Search it directly with `rg`; do not introduce a search index, vector database, or generated copy of the wiki.
-
-## Locate the wiki
-
-Resolve the checkout in this order:
-
-1. `$HOMELAB_WIKI_PATH`, when set.
-2. A sibling directory named `homelab-wiki` next to the current repository.
-3. `/home/panxiao81/homelab-wiki` when it exists.
-
-If no checkout is available, report that constraint. Do not silently skip the knowledge step, clone a repository, or create a replacement wiki without the user's authorization.
-
-Before using the wiki, read its `AGENTS.md` completely. For edits, also read `README.md` and `CONTRIBUTING.md` completely and follow any more specific instructions associated with the target page.
-
-## Gather context
-
-At the beginning of a homelab task:
-
-1. Derive search terms from the component name, service aliases, hostnames, Kubernetes resources, configuration keys, error text, and task intent.
-2. Use `rg -n -i` in the wiki to find candidate pages. Prefer several precise searches over reading the whole repository.
-3. Follow the wiki's task index, service index, architecture constraints, source records, and verification conflicts when they are relevant.
-4. Read the closest authoritative pages and their material links before making decisions. Also read the corresponding source repository README or runbook when changing an implementation.
-5. Distinguish documented design, declared configuration, deployment history, live verification, and work currently in progress. Do not present one as another.
-
-For questions about current project or service status, first obtain the maintainer's current-work and ticket context as required by the wiki, unless the conversation already provides that authorization and scope. Reading documentation does not authorize live-system inspection.
-
-Answer read-only questions from the evidence found. Include paths or links that let the user verify important claims, and state when evidence may be stale or conflicting.
-
-## Maintain knowledge after changes
-
-For any code, configuration, infrastructure, or operational change, perform a documentation-impact check before declaring the task complete.
-
-Update the wiki in the same task when the change affects durable knowledge such as:
-
-- service purpose, lifecycle, entry point, authentication, permissions, dependencies, or first-use path;
-- architecture boundaries or accepted constraints;
-- deployment ownership or persistent operating behavior;
-- troubleshooting, recovery, verification, or maintenance procedures;
-- the addition, replacement, or retirement of a service.
-
-Keep one-time progress, implementation narration, and release-by-release history in commits, PRs, or tickets. Do not copy them into the wiki unless they change a durable stage summary. Implementation-specific parameters may remain in the source repository README or runbook when the wiki convention says to link rather than duplicate them.
-
-When editing:
-
-1. Inspect both the source-repository diff and the wiki working tree before writing. Preserve unrelated user changes in both repositories.
-2. Update the page closest to the fact first, then only the navigation, indexes, constraints, or verification records that the wiki rules require.
-3. Preserve evidence metadata. Never advance `last_verified` without performing the stated live verification; ordinary review may update only fields permitted by the wiki.
-4. Link related source commits, PRs, or paths when available. Clearly mark uncommitted sources and unfinished cross-repository synchronization.
-5. Record conflicts rather than resolving them by assumption. Ask before live inspection or before choosing among materially conflicting current-state claims.
-6. Keep credentials, tokens, private keys, Terraform state, secret values, and sensitive command output out of documentation. Never read or copy known sensitive files merely to improve the wiki.
-
-Wiki edits are a separate repository change. Do not commit, push, open a PR, or modify a live system unless the user has authorized that action.
-
-## Verify and report
-
-After editing the wiki, run from its root:
-
-```bash
-python3 scripts/check_docs.py
-git diff --check
-```
-
-If the checker itself changed, also run:
-
-```bash
-python3 -m unittest discover -s tests -v
-```
-
-In the final response, report source-repository changes and wiki changes separately, including validation performed and anything still awaiting verification or cross-repository linkage. If no wiki update was needed, state the concrete reason; do not merely say that documentation was unaffected.

cmd/main.go

@@ -20,6 +20,7 @@ import (
 	"sigs.k8s.io/controller-runtime/pkg/webhook"
 
 	executionv1alpha1 "git.ddupan.top/panxiao81/ayatori/api/execution/v1alpha1"
+	"git.ddupan.top/panxiao81/ayatori/internal/controller"
 	// +kubebuilder:scaffold:imports
 )
 
@@ -165,6 +166,13 @@ func main() {
 		os.Exit(1)
 	}
 
+	if err := (&controller.JobReconciler{
+		Client: mgr.GetClient(),
+	}).SetupWithManager(mgr); err != nil {
+		setupLog.Error(err, "Failed to create controller", "controller", "Job")
+		os.Exit(1)
+	}
+
 	// +kubebuilder:scaffold:builder
 
 	if err := mgr.AddHealthzCheck("healthz", healthz.Ping); err != nil {

config/rbac/role.yaml

@@ -1,11 +1,58 @@
+---
 apiVersion: rbac.authorization.k8s.io/v1
 kind: ClusterRole
 metadata:
-  labels:
-    app.kubernetes.io/name: ayatori
-    app.kubernetes.io/managed-by: kustomize
   name: manager-role
 rules:
-- apiGroups: [""]
-  resources: ["pods"]
-  verbs: ["get", "list", "watch"]
+- apiGroups:
+  - ""
+  resources:
+  - namespaces
+  - serviceaccounts
+  verbs:
+  - get
+  - list
+  - watch
+- apiGroups:
+  - batch
+  resources:
+  - jobs
+  verbs:
+  - create
+  - delete
+  - get
+  - list
+  - watch
+- apiGroups:
+  - execution.ayatori.ddupan.top
+  resources:
+  - jobclasses
+  - kubernetesexecutionparameters
+  verbs:
+  - get
+  - list
+  - watch
+- apiGroups:
+  - execution.ayatori.ddupan.top
+  resources:
+  - jobs
+  verbs:
+  - get
+  - list
+  - patch
+  - update
+  - watch
+- apiGroups:
+  - execution.ayatori.ddupan.top
+  resources:
+  - jobs/finalizers
+  verbs:
+  - update
+- apiGroups:
+  - execution.ayatori.ddupan.top
+  resources:
+  - jobs/status
+  verbs:
+  - get
+  - patch
+  - update

docs/roadmap.md

@@ -51,10 +51,9 @@
 - 用户集群只暴露 worker node，控制面完全由平台托管。
 - 本节记录候选实现边界，不构成路线图承诺。
 
-## 后续 Compute 验收场景
+## 首个业务里程碑
 
-Database 等首批资源优先落地。Compute 开始实施后，以 Laptop Rebuild Readiness 验证节点
-生命周期与恢复能力；该场景不作为首批 Database、LoadBalancer 或 Bucket 的交付前置条件：
+完成 Laptop Rebuild Readiness：
 
 1. 临时节点加入。
 2. laptop 上的 workload 被重建、迁移或形成可执行人工任务。

internal/adapter/kubernetes/job.go

@@ -0,0 +1,132 @@
+package kubernetes
+
+import (
+	"fmt"
+
+	executionv1alpha1 "git.ddupan.top/panxiao81/ayatori/api/execution/v1alpha1"
+	batchv1 "k8s.io/api/batch/v1"
+	corev1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/runtime/schema"
+)
+
+const (
+	ControllerName = "execution.ayatori.ddupan.top/kubernetes"
+	ReferenceType  = "Job"
+	JobUIDLabel    = "execution.ayatori.ddupan.top/job-uid"
+)
+
+var ayatoriJobGVK = schema.GroupVersionKind{
+	Group:   executionv1alpha1.GroupVersion.Group,
+	Version: executionv1alpha1.GroupVersion.Version,
+	Kind:    "Job",
+}
+
+// BuildJob translates the stable execution API into the Kubernetes adapter's
+// backend object. It intentionally does not accept or expose a PodSpec.
+func BuildJob(
+	job *executionv1alpha1.Job,
+	parameters *executionv1alpha1.KubernetesExecutionParameters,
+	resources executionv1alpha1.ExecutionResourceRequirements,
+) *batchv1.Job {
+	backoffLimit := int32(0)
+	controller := true
+	blockOwnerDeletion := true
+
+	//nolint:modernize // ObjectMeta is promoted through embedded TypeMeta; embedlit produces invalid Go here.
+	return &batchv1.Job{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      job.Name,
+			Namespace: job.Namespace,
+			Labels: map[string]string{
+				JobUIDLabel: string(job.UID),
+			},
+			OwnerReferences: []metav1.OwnerReference{{
+				APIVersion:         ayatoriJobGVK.GroupVersion().String(),
+				Kind:               ayatoriJobGVK.Kind,
+				Name:               job.Name,
+				UID:                job.UID,
+				Controller:         &controller,
+				BlockOwnerDeletion: &blockOwnerDeletion,
+			}},
+		},
+		Spec: batchv1.JobSpec{
+			BackoffLimit: &backoffLimit,
+			Template: corev1.PodTemplateSpec{
+				ObjectMeta: metav1.ObjectMeta{Labels: map[string]string{JobUIDLabel: string(job.UID)}},
+				Spec: corev1.PodSpec{
+					RestartPolicy:      corev1.RestartPolicyNever,
+					ServiceAccountName: parameters.Spec.ServiceAccountName,
+					RuntimeClassName:   optionalString(parameters.Spec.RuntimeClassName),
+					NodeSelector:       parameters.Spec.Scheduling.NodeSelector,
+					Tolerations:        parameters.Spec.Scheduling.Tolerations,
+					SecurityContext:    parameters.Spec.PodSecurityContext,
+					ImagePullSecrets:   job.Spec.Task.ImagePullSecrets,
+					Containers: []corev1.Container{{
+						Name:            "task",
+						Image:           job.Spec.Task.Image,
+						ImagePullPolicy: parameters.Spec.ImagePullPolicy,
+						Command:         job.Spec.Task.Command,
+						Args:            job.Spec.Task.Args,
+						WorkingDir:      job.Spec.Task.WorkingDir,
+						Env:             environment(job.Spec.Task.Env),
+						Resources:       resourceRequirements(resources),
+					}},
+				},
+			},
+		},
+	}
+}
+
+func ValidateOwnership(owner *executionv1alpha1.Job, backend *batchv1.Job) error {
+	if backend.Labels[JobUIDLabel] != string(owner.UID) {
+		return fmt.Errorf("backend Job %s/%s is not owned by Ayatori Job UID %s", backend.Namespace, backend.Name, owner.UID)
+	}
+	return nil
+}
+
+func environment(values []executionv1alpha1.EnvVar) []corev1.EnvVar {
+	result := make([]corev1.EnvVar, 0, len(values))
+	for _, value := range values {
+		env := corev1.EnvVar{Name: value.Name}
+		if value.Value != nil {
+			env.Value = *value.Value
+		}
+		if value.ValueFrom != nil {
+			env.ValueFrom = &corev1.EnvVarSource{
+				SecretKeyRef:    value.ValueFrom.SecretKeyRef,
+				ConfigMapKeyRef: value.ValueFrom.ConfigMapKeyRef,
+			}
+		}
+		result = append(result, env)
+	}
+	return result
+}
+
+func resourceRequirements(resources executionv1alpha1.ExecutionResourceRequirements) corev1.ResourceRequirements {
+	return corev1.ResourceRequirements{
+		Requests: resourceList(resources.Requests),
+		Limits:   resourceList(resources.Limits),
+	}
+}
+
+func resourceList(values executionv1alpha1.ResourceValues) corev1.ResourceList {
+	result := corev1.ResourceList{}
+	if values.CPU != nil {
+		result[corev1.ResourceCPU] = values.CPU.DeepCopy()
+	}
+	if values.Memory != nil {
+		result[corev1.ResourceMemory] = values.Memory.DeepCopy()
+	}
+	if len(result) == 0 {
+		return nil
+	}
+	return result
+}
+
+func optionalString(value string) *string {
+	if value == "" {
+		return nil
+	}
+	return &value
+}

internal/adapter/kubernetes/job_test.go

@@ -0,0 +1,71 @@
+package kubernetes
+
+import (
+	"testing"
+
+	executionv1alpha1 "git.ddupan.top/panxiao81/ayatori/api/execution/v1alpha1"
+	corev1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/resource"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/types"
+)
+
+func TestBuildJob(t *testing.T) {
+	literal := "world"
+	cpuRequest := resource.MustParse("100m")
+	memoryLimit := resource.MustParse("128Mi")
+	//nolint:modernize // ObjectMeta is promoted through embedded TypeMeta; embedlit produces invalid Go here.
+	job := &executionv1alpha1.Job{
+		ObjectMeta: metav1.ObjectMeta{Name: "hello", Namespace: "ci", UID: types.UID("job-uid")},
+		Spec: executionv1alpha1.JobSpec{Task: executionv1alpha1.TaskSpec{
+			Image: "alpine:3.22", Command: []string{"echo"}, Args: []string{"hello"},
+			Env: []executionv1alpha1.EnvVar{
+				{Name: "TARGET", Value: &literal},
+				{Name: "TOKEN", ValueFrom: &executionv1alpha1.EnvVarSource{
+					//nolint:modernize // LocalObjectReference is an embedded Kubernetes API field.
+					SecretKeyRef: &corev1.SecretKeySelector{LocalObjectReference: corev1.LocalObjectReference{Name: "token"}, Key: "value"},
+				}},
+			},
+		}},
+	}
+	parameters := &executionv1alpha1.KubernetesExecutionParameters{Spec: executionv1alpha1.KubernetesExecutionParametersSpec{
+		ServiceAccountName: "runner", RuntimeClassName: "runc", ImagePullPolicy: corev1.PullIfNotPresent,
+		Scheduling: executionv1alpha1.KubernetesSchedulingParameters{NodeSelector: map[string]string{"role": "execution"}},
+	}}
+	resources := executionv1alpha1.ExecutionResourceRequirements{
+		Requests: executionv1alpha1.ResourceValues{CPU: &cpuRequest},
+		Limits:   executionv1alpha1.ResourceValues{Memory: &memoryLimit},
+	}
+
+	backend := BuildJob(job, parameters, resources)
+	pod := backend.Spec.Template.Spec
+	if backend.Spec.BackoffLimit == nil || *backend.Spec.BackoffLimit != 0 {
+		t.Fatalf("backoffLimit = %v, want 0", backend.Spec.BackoffLimit)
+	}
+	if pod.RestartPolicy != corev1.RestartPolicyNever || pod.ServiceAccountName != "runner" {
+		t.Fatalf("unexpected pod execution policy: %#v", pod)
+	}
+	if pod.RuntimeClassName == nil || *pod.RuntimeClassName != "runc" {
+		t.Fatalf("runtimeClassName = %v, want runc", pod.RuntimeClassName)
+	}
+	container := pod.Containers[0]
+	if container.Resources.Requests.Cpu().Cmp(cpuRequest) != 0 || container.Resources.Limits.Memory().Cmp(memoryLimit) != 0 {
+		t.Fatalf("resources were not mapped: %#v", container.Resources)
+	}
+	if container.Env[1].ValueFrom == nil || container.Env[1].ValueFrom.SecretKeyRef.Name != "token" {
+		t.Fatalf("secret reference was not preserved: %#v", container.Env[1])
+	}
+	if backend.Labels[JobUIDLabel] != "job-uid" || backend.OwnerReferences[0].UID != job.UID {
+		t.Fatalf("ownership identity was not preserved: %#v", backend.ObjectMeta)
+	}
+}
+
+func TestValidateOwnership(t *testing.T) {
+	job := &executionv1alpha1.Job{}
+	job.UID = types.UID("expected")
+	backend := BuildJob(job, &executionv1alpha1.KubernetesExecutionParameters{}, executionv1alpha1.ExecutionResourceRequirements{})
+	backend.Labels[JobUIDLabel] = "different"
+	if err := ValidateOwnership(job, backend); err == nil {
+		t.Fatal("ValidateOwnership() succeeded for a different Job UID")
+	}
+}

internal/controller/job_controller.go

@@ -0,0 +1,397 @@
+package controller
+
+import (
+	"context"
+	"fmt"
+	"slices"
+	"time"
+
+	executionv1alpha1 "git.ddupan.top/panxiao81/ayatori/api/execution/v1alpha1"
+	kubernetesadapter "git.ddupan.top/panxiao81/ayatori/internal/adapter/kubernetes"
+	batchv1 "k8s.io/api/batch/v1"
+	corev1 "k8s.io/api/core/v1"
+	apierrors "k8s.io/apimachinery/pkg/api/errors"
+	"k8s.io/apimachinery/pkg/api/meta"
+	"k8s.io/apimachinery/pkg/api/resource"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/labels"
+	"k8s.io/apimachinery/pkg/types"
+	ctrl "sigs.k8s.io/controller-runtime"
+	"sigs.k8s.io/controller-runtime/pkg/client"
+	"sigs.k8s.io/controller-runtime/pkg/log"
+)
+
+const (
+	jobFinalizer        = "execution.ayatori.ddupan.top/job-cleanup"
+	reasonResultUnknown = "ResultUnknown"
+)
+
+// JobReconciler executes Ayatori Jobs using supported adapters.
+type JobReconciler struct {
+	client.Client
+	Now func() time.Time
+}
+
+// +kubebuilder:rbac:groups=execution.ayatori.ddupan.top,resources=jobs,verbs=get;list;watch;update;patch
+// +kubebuilder:rbac:groups=execution.ayatori.ddupan.top,resources=jobs/status,verbs=get;update;patch
+// +kubebuilder:rbac:groups=execution.ayatori.ddupan.top,resources=jobs/finalizers,verbs=update
+// +kubebuilder:rbac:groups=execution.ayatori.ddupan.top,resources=jobclasses;kubernetesexecutionparameters,verbs=get;list;watch
+// +kubebuilder:rbac:groups=batch,resources=jobs,verbs=get;list;watch;create;delete
+// +kubebuilder:rbac:groups="",resources=namespaces;serviceaccounts,verbs=get;list;watch
+
+func (r *JobReconciler) Reconcile(ctx context.Context, request ctrl.Request) (ctrl.Result, error) {
+	logger := log.FromContext(ctx)
+	job := &executionv1alpha1.Job{}
+	if err := r.Get(ctx, request.NamespacedName, job); err != nil {
+		return ctrl.Result{}, client.IgnoreNotFound(err)
+	}
+
+	if !job.DeletionTimestamp.IsZero() {
+		return ctrl.Result{}, r.finalize(ctx, job)
+	}
+	if isTerminal(job) {
+		return ctrl.Result{}, nil
+	}
+	if job.Spec.DesiredState == executionv1alpha1.JobDesiredStateCancelled {
+		return ctrl.Result{}, r.cancel(ctx, job)
+	}
+
+	if !containsString(job.Finalizers, jobFinalizer) {
+		job.Finalizers = append(job.Finalizers, jobFinalizer)
+		if err := r.Update(ctx, job); err != nil {
+			return ctrl.Result{}, err
+		}
+		return ctrl.Result{}, nil
+	}
+	if job.Status.Execution != nil {
+		return ctrl.Result{}, r.observeExisting(ctx, job)
+	}
+
+	class, parameters, resources, waiting, err := r.resolve(ctx, job)
+	if err != nil {
+		return ctrl.Result{}, err
+	}
+	if waiting {
+		return ctrl.Result{RequeueAfter: 30 * time.Second}, nil
+	}
+
+	backend := &batchv1.Job{}
+	key := types.NamespacedName{Namespace: job.Namespace, Name: job.Name}
+	err = r.Get(ctx, key, backend)
+	if apierrors.IsNotFound(err) {
+		backend = kubernetesadapter.BuildJob(job, parameters, resources)
+		if err := r.Create(ctx, backend); err != nil {
+			return ctrl.Result{}, err
+		}
+		logger.Info("Created Kubernetes backend Job", "backend", key)
+		return ctrl.Result{}, r.markScheduled(ctx, job, class, parameters, resources, backend)
+	}
+	if err != nil {
+		return ctrl.Result{}, err
+	}
+	if err := kubernetesadapter.ValidateOwnership(job, backend); err != nil {
+		return ctrl.Result{}, r.setCondition(ctx, job, metav1.Condition{
+			Type: executionv1alpha1.JobConditionScheduled, Status: metav1.ConditionFalse,
+			Reason: "BackendConflict", Message: err.Error(),
+		})
+	}
+	if !conditionTrue(job.Status.Conditions, executionv1alpha1.JobConditionScheduled) {
+		return ctrl.Result{}, r.markScheduled(ctx, job, class, parameters, resources, backend)
+	}
+	return ctrl.Result{}, r.observe(ctx, job, backend)
+}
+
+func (r *JobReconciler) observeExisting(ctx context.Context, job *executionv1alpha1.Job) error {
+	if job.Status.Execution.Adapter != "kubernetes" {
+		return r.setCondition(ctx, job, metav1.Condition{
+			Type: executionv1alpha1.JobConditionSucceeded, Status: metav1.ConditionUnknown,
+			Reason: reasonResultUnknown, Message: fmt.Sprintf("adapter %q is not available", job.Status.Execution.Adapter),
+		})
+	}
+	backend := &batchv1.Job{}
+	key := types.NamespacedName{Namespace: job.Namespace, Name: job.Name}
+	if err := r.Get(ctx, key, backend); err != nil {
+		if apierrors.IsNotFound(err) {
+			return r.setCondition(ctx, job, metav1.Condition{
+				Type: executionv1alpha1.JobConditionSucceeded, Status: metav1.ConditionUnknown,
+				Reason: reasonResultUnknown, Message: "Kubernetes backend Job is missing",
+			})
+		}
+		return err
+	}
+	if err := kubernetesadapter.ValidateOwnership(job, backend); err != nil {
+		return r.setCondition(ctx, job, metav1.Condition{
+			Type: executionv1alpha1.JobConditionSucceeded, Status: metav1.ConditionUnknown,
+			Reason: reasonResultUnknown, Message: err.Error(),
+		})
+	}
+	return r.observe(ctx, job, backend)
+}
+
+func (r *JobReconciler) resolve(
+	ctx context.Context,
+	job *executionv1alpha1.Job,
+) (*executionv1alpha1.JobClass, *executionv1alpha1.KubernetesExecutionParameters, executionv1alpha1.ExecutionResourceRequirements, bool, error) {
+	if job.Spec.JobClassName == "" {
+		return nil, nil, executionv1alpha1.ExecutionResourceRequirements{}, true, r.reject(ctx, job, "NoDefaultJobClass", "spec.jobClassName is required in the first implementation slice")
+	}
+
+	class := &executionv1alpha1.JobClass{}
+	if err := r.Get(ctx, types.NamespacedName{Name: job.Spec.JobClassName}, class); err != nil {
+		if apierrors.IsNotFound(err) {
+			return nil, nil, executionv1alpha1.ExecutionResourceRequirements{}, true, r.reject(ctx, job, "JobClassNotFound", fmt.Sprintf("JobClass %q does not exist", job.Spec.JobClassName))
+		}
+		return nil, nil, executionv1alpha1.ExecutionResourceRequirements{}, false, err
+	}
+	if class.Spec.ControllerName != kubernetesadapter.ControllerName {
+		return nil, nil, executionv1alpha1.ExecutionResourceRequirements{}, true, r.reject(ctx, job, "UnsupportedController", fmt.Sprintf("controller %q is not supported", class.Spec.ControllerName))
+	}
+	ref := class.Spec.ParametersRef
+	if ref.Group != executionv1alpha1.GroupVersion.Group || ref.Kind != "KubernetesExecutionParameters" {
+		return nil, nil, executionv1alpha1.ExecutionResourceRequirements{}, true, r.reject(ctx, job, "InvalidParametersReference", "JobClass must reference KubernetesExecutionParameters")
+	}
+	if allowed, err := r.namespaceAllowed(ctx, job.Namespace, class.Spec.AllowedNamespaces); err != nil {
+		return nil, nil, executionv1alpha1.ExecutionResourceRequirements{}, false, err
+	} else if !allowed {
+		return nil, nil, executionv1alpha1.ExecutionResourceRequirements{}, true, r.reject(ctx, job, "NamespaceNotAllowed", fmt.Sprintf("namespace %q is not allowed by JobClass %q", job.Namespace, class.Name))
+	}
+
+	parameters := &executionv1alpha1.KubernetesExecutionParameters{}
+	if err := r.Get(ctx, types.NamespacedName{Name: ref.Name}, parameters); err != nil {
+		if apierrors.IsNotFound(err) {
+			return nil, nil, executionv1alpha1.ExecutionResourceRequirements{}, true, r.reject(ctx, job, "ParametersNotFound", fmt.Sprintf("KubernetesExecutionParameters %q does not exist", ref.Name))
+		}
+		return nil, nil, executionv1alpha1.ExecutionResourceRequirements{}, false, err
+	}
+	serviceAccount := &corev1.ServiceAccount{}
+	if err := r.Get(ctx, types.NamespacedName{Namespace: job.Namespace, Name: parameters.Spec.ServiceAccountName}, serviceAccount); err != nil {
+		if apierrors.IsNotFound(err) {
+			return nil, nil, executionv1alpha1.ExecutionResourceRequirements{}, true, r.reject(ctx, job, "ServiceAccountNotFound", fmt.Sprintf("ServiceAccount %q does not exist", parameters.Spec.ServiceAccountName))
+		}
+		return nil, nil, executionv1alpha1.ExecutionResourceRequirements{}, false, err
+	}
+
+	resources := applyResourceDefaults(job.Spec.Resources, class.Spec.Resources.Defaults)
+	if err := validateResources(resources); err != nil {
+		return nil, nil, executionv1alpha1.ExecutionResourceRequirements{}, true, r.reject(ctx, job, "InvalidResources", err.Error())
+	}
+	if err := r.accept(ctx, job, class, parameters, resources); err != nil {
+		return nil, nil, executionv1alpha1.ExecutionResourceRequirements{}, false, err
+	}
+	return class, parameters, resources, false, nil
+}
+
+func (r *JobReconciler) namespaceAllowed(ctx context.Context, namespace string, selector *metav1.LabelSelector) (bool, error) {
+	if selector == nil {
+		return true, nil
+	}
+	ns := &corev1.Namespace{}
+	if err := r.Get(ctx, types.NamespacedName{Name: namespace}, ns); err != nil {
+		return false, err
+	}
+	compiled, err := metav1.LabelSelectorAsSelector(selector)
+	if err != nil {
+		return false, err
+	}
+	return compiled.Matches(labels.Set(ns.Labels)), nil
+}
+
+func (r *JobReconciler) accept(ctx context.Context, job *executionv1alpha1.Job, class *executionv1alpha1.JobClass, parameters *executionv1alpha1.KubernetesExecutionParameters, resources executionv1alpha1.ExecutionResourceRequirements) error {
+	job.Status.ResolvedJobClass = &executionv1alpha1.ResolvedJobClassReference{
+		Name: class.Name, UID: class.UID, ControllerName: class.Spec.ControllerName,
+		ParametersRef: executionv1alpha1.ParametersReference{
+			Group: class.Spec.ParametersRef.Group, Kind: class.Spec.ParametersRef.Kind,
+			Name: parameters.Name, UID: parameters.UID,
+		},
+	}
+	job.Status.EffectiveResources = resources
+	return r.setCondition(ctx, job, metav1.Condition{
+		Type: executionv1alpha1.JobConditionAccepted, Status: metav1.ConditionTrue,
+		Reason: "Accepted", Message: fmt.Sprintf("JobClass %q accepted", class.Name),
+	})
+}
+
+func (r *JobReconciler) reject(ctx context.Context, job *executionv1alpha1.Job, reason, message string) error {
+	return r.setCondition(ctx, job, metav1.Condition{
+		Type: executionv1alpha1.JobConditionAccepted, Status: metav1.ConditionFalse,
+		Reason: reason, Message: message,
+	})
+}
+
+func (r *JobReconciler) markScheduled(ctx context.Context, job *executionv1alpha1.Job, class *executionv1alpha1.JobClass, parameters *executionv1alpha1.KubernetesExecutionParameters, resources executionv1alpha1.ExecutionResourceRequirements, backend *batchv1.Job) error {
+	job.Status.ResolvedJobClass = &executionv1alpha1.ResolvedJobClassReference{
+		Name: class.Name, UID: class.UID, ControllerName: class.Spec.ControllerName,
+		ParametersRef: executionv1alpha1.ParametersReference{Group: class.Spec.ParametersRef.Group, Kind: class.Spec.ParametersRef.Kind, Name: parameters.Name, UID: parameters.UID},
+	}
+	job.Status.EffectiveResources = resources
+	job.Status.Execution = &executionv1alpha1.ExecutionStatus{
+		Adapter:    "kubernetes",
+		References: []executionv1alpha1.ExecutionReference{{Type: kubernetesadapter.ReferenceType, ID: string(backend.UID)}},
+	}
+	meta.SetStatusCondition(&job.Status.Conditions, condition(job, executionv1alpha1.JobConditionAccepted, metav1.ConditionTrue, "Accepted", "Job accepted"))
+	meta.SetStatusCondition(&job.Status.Conditions, condition(job, executionv1alpha1.JobConditionScheduled, metav1.ConditionTrue, "BackendCreated", "Kubernetes Job created"))
+	meta.SetStatusCondition(&job.Status.Conditions, condition(job, executionv1alpha1.JobConditionSucceeded, metav1.ConditionUnknown, "Pending", "Waiting for task to start"))
+	job.Status.ObservedGeneration = job.Generation
+	return r.Status().Update(ctx, job)
+}
+
+func (r *JobReconciler) observe(ctx context.Context, job *executionv1alpha1.Job, backend *batchv1.Job) error {
+	if job.Status.StartTime == nil && backend.Status.StartTime != nil {
+		job.Status.StartTime = backend.Status.StartTime.DeepCopy()
+	}
+	for _, backendCondition := range backend.Status.Conditions {
+		switch {
+		case backendCondition.Type == batchv1.JobComplete && backendCondition.Status == corev1.ConditionTrue:
+			completion := backend.Status.CompletionTime
+			if completion == nil {
+				now := metav1.NewTime(r.now())
+				completion = &now
+			}
+			job.Status.CompletionTime = completion.DeepCopy()
+			job.Status.Result = &executionv1alpha1.JobResult{Reason: "Completed"}
+			return r.setCondition(ctx, job, metav1.Condition{Type: executionv1alpha1.JobConditionSucceeded, Status: metav1.ConditionTrue, Reason: "Completed", Message: backendCondition.Message})
+		case backendCondition.Type == batchv1.JobFailed && backendCondition.Status == corev1.ConditionTrue:
+			completion := metav1.NewTime(r.now())
+			job.Status.CompletionTime = &completion
+			job.Status.Result = &executionv1alpha1.JobResult{Reason: "ProcessFailed"}
+			return r.setCondition(ctx, job, metav1.Condition{Type: executionv1alpha1.JobConditionSucceeded, Status: metav1.ConditionFalse, Reason: "ProcessFailed", Message: backendCondition.Message})
+		}
+	}
+	reason := "Pending"
+	message := "Waiting for task to start"
+	if backend.Status.StartTime != nil || backend.Status.Active > 0 {
+		reason = "Running"
+		message = "Task is running"
+	}
+	return r.setCondition(ctx, job, metav1.Condition{Type: executionv1alpha1.JobConditionSucceeded, Status: metav1.ConditionUnknown, Reason: reason, Message: message})
+}
+
+func (r *JobReconciler) cancel(ctx context.Context, job *executionv1alpha1.Job) error {
+	backend := &batchv1.Job{}
+	key := types.NamespacedName{Namespace: job.Namespace, Name: job.Name}
+	err := r.Get(ctx, key, backend)
+	if err == nil {
+		if err := kubernetesadapter.ValidateOwnership(job, backend); err != nil {
+			return err
+		}
+		for _, backendCondition := range backend.Status.Conditions {
+			if (backendCondition.Type == batchv1.JobComplete || backendCondition.Type == batchv1.JobFailed) &&
+				backendCondition.Status == corev1.ConditionTrue {
+				return r.observe(ctx, job, backend)
+			}
+		}
+		if err := r.Delete(ctx, backend, client.PropagationPolicy(metav1.DeletePropagationBackground)); err != nil && !apierrors.IsNotFound(err) {
+			return err
+		}
+		return nil
+	}
+	if !apierrors.IsNotFound(err) {
+		return err
+	}
+	now := metav1.NewTime(r.now())
+	job.Status.CompletionTime = &now
+	job.Status.Result = &executionv1alpha1.JobResult{Reason: "Cancelled"}
+	return r.setCondition(ctx, job, metav1.Condition{Type: executionv1alpha1.JobConditionSucceeded, Status: metav1.ConditionFalse, Reason: "Cancelled", Message: "Execution cancelled"})
+}
+
+func (r *JobReconciler) finalize(ctx context.Context, job *executionv1alpha1.Job) error {
+	if !containsString(job.Finalizers, jobFinalizer) {
+		return nil
+	}
+	backend := &batchv1.Job{}
+	key := types.NamespacedName{Namespace: job.Namespace, Name: job.Name}
+	if err := r.Get(ctx, key, backend); err == nil {
+		if err := kubernetesadapter.ValidateOwnership(job, backend); err != nil {
+			return err
+		}
+		if err := r.Delete(ctx, backend, client.PropagationPolicy(metav1.DeletePropagationBackground)); err != nil && !apierrors.IsNotFound(err) {
+			return err
+		}
+		return nil
+	} else if !apierrors.IsNotFound(err) {
+		return err
+	}
+	job.Finalizers = removeString(job.Finalizers, jobFinalizer)
+	return r.Update(ctx, job)
+}
+
+func (r *JobReconciler) setCondition(ctx context.Context, job *executionv1alpha1.Job, next metav1.Condition) error {
+	meta.SetStatusCondition(&job.Status.Conditions, condition(job, next.Type, next.Status, next.Reason, next.Message))
+	job.Status.ObservedGeneration = job.Generation
+	return r.Status().Update(ctx, job)
+}
+
+func condition(job *executionv1alpha1.Job, conditionType string, status metav1.ConditionStatus, reason, message string) metav1.Condition {
+	return metav1.Condition{Type: conditionType, Status: status, Reason: reason, Message: message, ObservedGeneration: job.Generation}
+}
+
+func conditionTrue(conditions []metav1.Condition, conditionType string) bool {
+	current := meta.FindStatusCondition(conditions, conditionType)
+	return current != nil && current.Status == metav1.ConditionTrue
+}
+
+func isTerminal(job *executionv1alpha1.Job) bool {
+	current := meta.FindStatusCondition(job.Status.Conditions, executionv1alpha1.JobConditionSucceeded)
+	return current != nil && (current.Status == metav1.ConditionTrue || current.Status == metav1.ConditionFalse)
+}
+
+func applyResourceDefaults(requested, defaults executionv1alpha1.ExecutionResourceRequirements) executionv1alpha1.ExecutionResourceRequirements {
+	result := requested.DeepCopy()
+	if result.Requests.CPU == nil && defaults.Requests.CPU != nil {
+		result.Requests.CPU = copyQuantity(defaults.Requests.CPU)
+	}
+	if result.Requests.Memory == nil && defaults.Requests.Memory != nil {
+		result.Requests.Memory = copyQuantity(defaults.Requests.Memory)
+	}
+	if result.Limits.CPU == nil && defaults.Limits.CPU != nil {
+		result.Limits.CPU = copyQuantity(defaults.Limits.CPU)
+	}
+	if result.Limits.Memory == nil && defaults.Limits.Memory != nil {
+		result.Limits.Memory = copyQuantity(defaults.Limits.Memory)
+	}
+	return *result
+}
+
+func copyQuantity(value *resource.Quantity) *resource.Quantity {
+	copy := value.DeepCopy()
+	return &copy
+}
+
+func validateResources(resources executionv1alpha1.ExecutionResourceRequirements) error {
+	if resources.Requests.CPU != nil && resources.Limits.CPU != nil && resources.Requests.CPU.Cmp(*resources.Limits.CPU) > 0 {
+		return fmt.Errorf("CPU request must not exceed limit")
+	}
+	if resources.Requests.Memory != nil && resources.Limits.Memory != nil && resources.Requests.Memory.Cmp(*resources.Limits.Memory) > 0 {
+		return fmt.Errorf("memory request must not exceed limit")
+	}
+	return nil
+}
+
+func containsString(values []string, target string) bool {
+	return slices.Contains(values, target)
+}
+
+func removeString(values []string, target string) []string {
+	result := values[:0]
+	for _, value := range values {
+		if value != target {
+			result = append(result, value)
+		}
+	}
+	return result
+}
+
+func (r *JobReconciler) now() time.Time {
+	if r.Now != nil {
+		return r.Now()
+	}
+	return time.Now()
+}
+
+func (r *JobReconciler) SetupWithManager(manager ctrl.Manager) error {
+	return ctrl.NewControllerManagedBy(manager).
+		For(&executionv1alpha1.Job{}).
+		Owns(&batchv1.Job{}).
+		Named("execution-job").
+		Complete(r)
+}

internal/controller/job_controller_integration_test.go

@@ -0,0 +1,202 @@
+package controller
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+
+	executionv1alpha1 "git.ddupan.top/panxiao81/ayatori/api/execution/v1alpha1"
+	kubernetesadapter "git.ddupan.top/panxiao81/ayatori/internal/adapter/kubernetes"
+	batchv1 "k8s.io/api/batch/v1"
+	corev1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/meta"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/types"
+	ctrl "sigs.k8s.io/controller-runtime"
+	"sigs.k8s.io/controller-runtime/pkg/client"
+	"sigs.k8s.io/controller-runtime/pkg/envtest"
+	metricsserver "sigs.k8s.io/controller-runtime/pkg/metrics/server"
+)
+
+const (
+	integrationNamespace = "controller-integration"
+	integrationClass     = "integration"
+)
+
+//nolint:modernize // Kubernetes API structs expose ObjectMeta through embedded TypeMeta fields.
+func TestJobControllerIntegration(t *testing.T) {
+	if os.Getenv("KUBEBUILDER_ASSETS") == "" {
+		t.Skip("KUBEBUILDER_ASSETS is unset; run make test to execute controller integration tests")
+	}
+
+	scheme := runtime.NewScheme()
+	for _, addToScheme := range []func(*runtime.Scheme) error{
+		corev1.AddToScheme,
+		batchv1.AddToScheme,
+		executionv1alpha1.AddToScheme,
+	} {
+		if err := addToScheme(scheme); err != nil {
+			t.Fatal(err)
+		}
+	}
+
+	crdPath, err := filepath.Abs("../../config/crd/bases")
+	if err != nil {
+		t.Fatal(err)
+	}
+	environment := &envtest.Environment{CRDDirectoryPaths: []string{crdPath}}
+	config, err := environment.Start()
+	if err != nil {
+		t.Fatalf("start envtest: %v", err)
+	}
+	t.Cleanup(func() {
+		if err := environment.Stop(); err != nil {
+			t.Errorf("stop envtest: %v", err)
+		}
+	})
+
+	manager, err := ctrl.NewManager(config, ctrl.Options{
+		Scheme:  scheme,
+		Metrics: metricsserver.Options{BindAddress: "0"},
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := (&JobReconciler{Client: manager.GetClient()}).SetupWithManager(manager); err != nil {
+		t.Fatal(err)
+	}
+
+	managerContext, cancelManager := context.WithCancel(context.Background())
+	t.Cleanup(cancelManager)
+	managerErrors := make(chan error, 1)
+	go func() {
+		managerErrors <- manager.Start(managerContext)
+	}()
+	if !manager.GetCache().WaitForCacheSync(managerContext) {
+		t.Fatal("manager cache did not synchronize")
+	}
+
+	directClient, err := client.New(config, client.Options{Scheme: scheme})
+	if err != nil {
+		t.Fatal(err)
+	}
+	ctx := context.Background()
+	objects := []client.Object{
+		&corev1.Namespace{ObjectMeta: metav1.ObjectMeta{Name: integrationNamespace, Labels: map[string]string{testLabelKey: testLabelEnabled}}},
+		&corev1.ServiceAccount{ObjectMeta: metav1.ObjectMeta{Name: testSAName, Namespace: integrationNamespace}},
+		&executionv1alpha1.KubernetesExecutionParameters{
+			ObjectMeta: metav1.ObjectMeta{Name: integrationClass},
+			Spec: executionv1alpha1.KubernetesExecutionParametersSpec{
+				ServiceAccountName: testSAName,
+				ImagePullPolicy:    corev1.PullIfNotPresent,
+			},
+		},
+		&executionv1alpha1.JobClass{
+			ObjectMeta: metav1.ObjectMeta{Name: integrationClass},
+			Spec: executionv1alpha1.JobClassSpec{
+				ControllerName: kubernetesadapter.ControllerName,
+				ParametersRef: executionv1alpha1.ParametersReference{
+					Group: executionv1alpha1.GroupVersion.Group,
+					Kind:  "KubernetesExecutionParameters",
+					Name:  integrationClass,
+				},
+				AllowedNamespaces: &metav1.LabelSelector{MatchLabels: map[string]string{testLabelKey: testLabelEnabled}},
+			},
+		},
+	}
+	for _, object := range objects {
+		if err := directClient.Create(ctx, object); err != nil {
+			t.Fatalf("create %T: %v", object, err)
+		}
+	}
+
+	job := &executionv1alpha1.Job{
+		ObjectMeta: metav1.ObjectMeta{Name: testJobName, Namespace: integrationNamespace},
+		Spec: executionv1alpha1.JobSpec{
+			JobClassName: integrationClass,
+			DesiredState: executionv1alpha1.JobDesiredStateRunning,
+			Task:         executionv1alpha1.TaskSpec{Image: "alpine:3.22", Command: []string{"true"}},
+		},
+	}
+	if err := directClient.Create(ctx, job); err != nil {
+		t.Fatal(err)
+	}
+
+	backend := &batchv1.Job{}
+	eventually(t, 10*time.Second, func() (bool, error) {
+		err := directClient.Get(ctx, types.NamespacedName{Namespace: job.Namespace, Name: job.Name}, backend)
+		return err == nil, client.IgnoreNotFound(err)
+	})
+	if backend.Labels[kubernetesadapter.JobUIDLabel] != string(job.UID) {
+		t.Fatalf("backend identity label = %q, want %q", backend.Labels[kubernetesadapter.JobUIDLabel], job.UID)
+	}
+
+	eventually(t, 10*time.Second, func() (bool, error) {
+		if err := directClient.Get(ctx, types.NamespacedName{Namespace: job.Namespace, Name: job.Name}, job); err != nil {
+			return false, err
+		}
+		return conditionStatus(job, executionv1alpha1.JobConditionScheduled) == metav1.ConditionTrue, nil
+	})
+
+	completed := metav1.Now()
+	if err := directClient.Get(ctx, types.NamespacedName{Namespace: job.Namespace, Name: job.Name}, backend); err != nil {
+		t.Fatal(err)
+	}
+	backend.Status.StartTime = &completed
+	backend.Status.CompletionTime = &completed
+	backend.Status.Conditions = []batchv1.JobCondition{
+		{Type: batchv1.JobSuccessCriteriaMet, Status: corev1.ConditionTrue, Reason: "CompletionsReached"},
+		{Type: batchv1.JobComplete, Status: corev1.ConditionTrue, Reason: "Completed"},
+	}
+	if err := directClient.Status().Update(ctx, backend); err != nil {
+		t.Fatal(err)
+	}
+
+	eventually(t, 10*time.Second, func() (bool, error) {
+		if err := directClient.Get(ctx, types.NamespacedName{Namespace: job.Namespace, Name: job.Name}, job); err != nil {
+			return false, err
+		}
+		return conditionStatus(job, executionv1alpha1.JobConditionSucceeded) == metav1.ConditionTrue, nil
+	})
+	if job.Status.StartTime == nil || job.Status.CompletionTime == nil || job.Status.Execution == nil {
+		t.Fatalf("controller did not persist execution status: %#v", job.Status)
+	}
+
+	cancelManager()
+	select {
+	case err := <-managerErrors:
+		if err != nil {
+			t.Fatalf("manager stopped with error: %v", err)
+		}
+	case <-time.After(5 * time.Second):
+		t.Fatal("manager did not stop")
+	}
+}
+
+func eventually(t *testing.T, timeout time.Duration, check func() (bool, error)) {
+	t.Helper()
+	deadline := time.Now().Add(timeout)
+	for time.Now().Before(deadline) {
+		ready, err := check()
+		if err != nil {
+			t.Fatal(err)
+		}
+		if ready {
+			return
+		}
+		time.Sleep(100 * time.Millisecond)
+	}
+	t.Fatal(fmt.Errorf("condition was not met within %s", timeout))
+}
+
+func conditionStatus(job *executionv1alpha1.Job, conditionType string) metav1.ConditionStatus {
+	condition := meta.FindStatusCondition(job.Status.Conditions, conditionType)
+	if condition == nil {
+		return metav1.ConditionUnknown
+	}
+	return condition.Status
+}

internal/controller/job_controller_test.go

@@ -0,0 +1,245 @@
+package controller
+
+import (
+	"context"
+	"testing"
+	"time"
+
+	executionv1alpha1 "git.ddupan.top/panxiao81/ayatori/api/execution/v1alpha1"
+	kubernetesadapter "git.ddupan.top/panxiao81/ayatori/internal/adapter/kubernetes"
+	batchv1 "k8s.io/api/batch/v1"
+	corev1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/meta"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/types"
+	ctrl "sigs.k8s.io/controller-runtime"
+	"sigs.k8s.io/controller-runtime/pkg/client"
+	"sigs.k8s.io/controller-runtime/pkg/client/fake"
+)
+
+const (
+	defaultClassName = "default"
+	testJobName      = "hello"
+	testSAName       = "runner"
+	testLabelKey     = "execution"
+	testLabelEnabled = "enabled"
+)
+
+//nolint:modernize // controller-runtime and Kubernetes API structs expose promoted embedded fields.
+func TestJobReconcilerKubernetesLifecycle(t *testing.T) {
+	ctx := context.Background()
+	now := time.Unix(1_700_000_000, 0)
+	reconciler, kubeClient := testReconciler(t, now, validObjects()...)
+	request := ctrl.Request{}
+	request.NamespacedName = types.NamespacedName{Namespace: "ci", Name: testJobName}
+
+	if _, err := reconciler.Reconcile(ctx, request); err != nil {
+		t.Fatalf("add finalizer: %v", err)
+	}
+	if _, err := reconciler.Reconcile(ctx, request); err != nil {
+		t.Fatalf("create backend: %v", err)
+	}
+
+	backend := &batchv1.Job{}
+	if err := kubeClient.Get(ctx, request.NamespacedName, backend); err != nil {
+		t.Fatalf("backend Job was not created: %v", err)
+	}
+	if backend.Labels[kubernetesadapter.JobUIDLabel] != "ayatori-job-uid" {
+		t.Fatalf("backend UID label = %q", backend.Labels[kubernetesadapter.JobUIDLabel])
+	}
+
+	job := getJob(t, ctx, kubeClient, request.NamespacedName)
+	if !conditionIs(job, executionv1alpha1.JobConditionAccepted, metav1.ConditionTrue) ||
+		!conditionIs(job, executionv1alpha1.JobConditionScheduled, metav1.ConditionTrue) {
+		t.Fatalf("Job was not accepted and scheduled: %#v", job.Status.Conditions)
+	}
+
+	started := metav1.NewTime(now.Add(time.Minute))
+	backend.Status.StartTime = &started
+	backend.Status.Active = 1
+	if err := kubeClient.Status().Update(ctx, backend); err != nil {
+		t.Fatalf("set backend running: %v", err)
+	}
+	if _, err := reconciler.Reconcile(ctx, request); err != nil {
+		t.Fatalf("observe running backend: %v", err)
+	}
+	job = getJob(t, ctx, kubeClient, request.NamespacedName)
+	if job.Status.StartTime == nil || !conditionIs(job, executionv1alpha1.JobConditionSucceeded, metav1.ConditionUnknown) {
+		t.Fatalf("running state was not observed: %#v", job.Status)
+	}
+
+	completed := metav1.NewTime(now.Add(2 * time.Minute))
+	backend = &batchv1.Job{}
+	if err := kubeClient.Get(ctx, request.NamespacedName, backend); err != nil {
+		t.Fatal(err)
+	}
+	backend.Status.Active = 0
+	backend.Status.CompletionTime = &completed
+	backend.Status.Conditions = []batchv1.JobCondition{{Type: batchv1.JobComplete, Status: corev1.ConditionTrue, Reason: "Completed"}}
+	if err := kubeClient.Status().Update(ctx, backend); err != nil {
+		t.Fatalf("set backend complete: %v", err)
+	}
+	if _, err := reconciler.Reconcile(ctx, request); err != nil {
+		t.Fatalf("observe completed backend: %v", err)
+	}
+	job = getJob(t, ctx, kubeClient, request.NamespacedName)
+	if !conditionIs(job, executionv1alpha1.JobConditionSucceeded, metav1.ConditionTrue) || job.Status.CompletionTime == nil {
+		t.Fatalf("terminal state was not observed: %#v", job.Status)
+	}
+}
+
+//nolint:modernize // controller-runtime Request exposes NamespacedName as a promoted embedded field.
+func TestJobReconcilerRejectsMissingClass(t *testing.T) {
+	ctx := context.Background()
+	job := validObjects()[3].(*executionv1alpha1.Job).DeepCopy()
+	job.Spec.JobClassName = "missing"
+	reconciler, kubeClient := testReconciler(t, time.Now(), validObjects()[0], validObjects()[1], job)
+	request := ctrl.Request{}
+	request.NamespacedName = types.NamespacedName{Namespace: job.Namespace, Name: job.Name}
+
+	if _, err := reconciler.Reconcile(ctx, request); err != nil {
+		t.Fatal(err)
+	}
+	result, err := reconciler.Reconcile(ctx, request)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if result.RequeueAfter == 0 {
+		t.Fatal("missing JobClass did not schedule a retry")
+	}
+	stored := getJob(t, ctx, kubeClient, request.NamespacedName)
+	accepted := meta.FindStatusCondition(stored.Status.Conditions, executionv1alpha1.JobConditionAccepted)
+	if accepted == nil || accepted.Status != metav1.ConditionFalse || accepted.Reason != "JobClassNotFound" {
+		t.Fatalf("unexpected Accepted condition: %#v", accepted)
+	}
+}
+
+//nolint:modernize // controller-runtime Request exposes NamespacedName as a promoted embedded field.
+func TestJobReconcilerObservesExistingExecutionWithoutJobClass(t *testing.T) {
+	ctx := context.Background()
+	now := time.Unix(1_700_000_000, 0)
+	job := validObjects()[3].(*executionv1alpha1.Job).DeepCopy()
+	job.Finalizers = []string{jobFinalizer}
+	job.Status.Execution = &executionv1alpha1.ExecutionStatus{Adapter: "kubernetes"}
+	backend := kubernetesadapter.BuildJob(job, validObjects()[4].(*executionv1alpha1.KubernetesExecutionParameters), executionv1alpha1.ExecutionResourceRequirements{})
+	backend.Status.StartTime = &metav1.Time{Time: now}
+	backend.Status.Active = 1
+	reconciler, kubeClient := testReconciler(t, now, job, backend)
+	request := ctrl.Request{NamespacedName: types.NamespacedName{Namespace: job.Namespace, Name: job.Name}}
+
+	if _, err := reconciler.Reconcile(ctx, request); err != nil {
+		t.Fatal(err)
+	}
+	stored := getJob(t, ctx, kubeClient, request.NamespacedName)
+	if stored.Status.StartTime == nil || !conditionIs(stored, executionv1alpha1.JobConditionSucceeded, metav1.ConditionUnknown) {
+		t.Fatalf("existing execution was not observed without its JobClass: %#v", stored.Status)
+	}
+}
+
+//nolint:modernize // controller-runtime Request exposes NamespacedName as a promoted embedded field.
+func TestJobReconcilerCancelsBeforeScheduling(t *testing.T) {
+	ctx := context.Background()
+	job := validObjects()[3].(*executionv1alpha1.Job).DeepCopy()
+	job.Spec.DesiredState = executionv1alpha1.JobDesiredStateCancelled
+	reconciler, kubeClient := testReconciler(t, time.Unix(1_700_000_000, 0), job)
+	request := ctrl.Request{NamespacedName: types.NamespacedName{Namespace: job.Namespace, Name: job.Name}}
+
+	if _, err := reconciler.Reconcile(ctx, request); err != nil {
+		t.Fatal(err)
+	}
+	stored := getJob(t, ctx, kubeClient, request.NamespacedName)
+	condition := meta.FindStatusCondition(stored.Status.Conditions, executionv1alpha1.JobConditionSucceeded)
+	if condition == nil || condition.Status != metav1.ConditionFalse || condition.Reason != "Cancelled" {
+		t.Fatalf("unexpected cancellation condition: %#v", condition)
+	}
+	if stored.Status.CompletionTime == nil {
+		t.Fatal("cancelled Job has no completionTime")
+	}
+}
+
+//nolint:modernize // controller-runtime Request exposes NamespacedName as a promoted embedded field.
+func TestJobReconcilerKeepsConfirmedSuccessDuringCancellation(t *testing.T) {
+	ctx := context.Background()
+	now := time.Unix(1_700_000_000, 0)
+	job := validObjects()[3].(*executionv1alpha1.Job).DeepCopy()
+	job.Spec.DesiredState = executionv1alpha1.JobDesiredStateCancelled
+	job.Finalizers = []string{jobFinalizer}
+	backend := kubernetesadapter.BuildJob(job, validObjects()[4].(*executionv1alpha1.KubernetesExecutionParameters), executionv1alpha1.ExecutionResourceRequirements{})
+	backend.Status.CompletionTime = &metav1.Time{Time: now}
+	backend.Status.Conditions = []batchv1.JobCondition{{Type: batchv1.JobComplete, Status: corev1.ConditionTrue}}
+	reconciler, kubeClient := testReconciler(t, now, job, backend)
+	request := ctrl.Request{NamespacedName: types.NamespacedName{Namespace: job.Namespace, Name: job.Name}}
+
+	if _, err := reconciler.Reconcile(ctx, request); err != nil {
+		t.Fatal(err)
+	}
+	stored := getJob(t, ctx, kubeClient, request.NamespacedName)
+	if !conditionIs(stored, executionv1alpha1.JobConditionSucceeded, metav1.ConditionTrue) {
+		t.Fatalf("confirmed success was overwritten by cancellation: %#v", stored.Status.Conditions)
+	}
+	if err := kubeClient.Get(ctx, request.NamespacedName, &batchv1.Job{}); err != nil {
+		t.Fatalf("successful backend was deleted: %v", err)
+	}
+}
+
+//nolint:modernize // Kubernetes API structs expose ObjectMeta through embedded TypeMeta fields.
+func validObjects() []client.Object {
+	return []client.Object{
+		&corev1.Namespace{ObjectMeta: metav1.ObjectMeta{Name: "ci", Labels: map[string]string{testLabelKey: testLabelEnabled}}},
+		&corev1.ServiceAccount{ObjectMeta: metav1.ObjectMeta{Name: testSAName, Namespace: "ci"}},
+		&executionv1alpha1.JobClass{
+			ObjectMeta: metav1.ObjectMeta{Name: defaultClassName, UID: types.UID("class-uid")},
+			Spec: executionv1alpha1.JobClassSpec{
+				ControllerName:    kubernetesadapter.ControllerName,
+				ParametersRef:     executionv1alpha1.ParametersReference{Group: executionv1alpha1.GroupVersion.Group, Kind: "KubernetesExecutionParameters", Name: defaultClassName},
+				AllowedNamespaces: &metav1.LabelSelector{MatchLabels: map[string]string{testLabelKey: testLabelEnabled}},
+			},
+		},
+		&executionv1alpha1.Job{
+			ObjectMeta: metav1.ObjectMeta{Name: testJobName, Namespace: "ci", UID: types.UID("ayatori-job-uid")},
+			Spec: executionv1alpha1.JobSpec{
+				JobClassName: defaultClassName, DesiredState: executionv1alpha1.JobDesiredStateRunning,
+				Task: executionv1alpha1.TaskSpec{Image: "alpine:3.22", Command: []string{"true"}},
+			},
+		},
+		&executionv1alpha1.KubernetesExecutionParameters{
+			ObjectMeta: metav1.ObjectMeta{Name: defaultClassName, UID: types.UID("parameters-uid")},
+			Spec:       executionv1alpha1.KubernetesExecutionParametersSpec{ServiceAccountName: testSAName, ImagePullPolicy: corev1.PullIfNotPresent},
+		},
+	}
+}
+
+func testReconciler(t *testing.T, now time.Time, objects ...client.Object) (*JobReconciler, client.Client) {
+	t.Helper()
+	scheme := runtime.NewScheme()
+	if err := corev1.AddToScheme(scheme); err != nil {
+		t.Fatal(err)
+	}
+	if err := batchv1.AddToScheme(scheme); err != nil {
+		t.Fatal(err)
+	}
+	if err := executionv1alpha1.AddToScheme(scheme); err != nil {
+		t.Fatal(err)
+	}
+	kubeClient := fake.NewClientBuilder().
+		WithScheme(scheme).
+		WithStatusSubresource(&executionv1alpha1.Job{}, &batchv1.Job{}).
+		WithObjects(objects...).
+		Build()
+	return &JobReconciler{Client: kubeClient, Now: func() time.Time { return now }}, kubeClient
+}
+
+func getJob(t *testing.T, ctx context.Context, kubeClient client.Client, key types.NamespacedName) *executionv1alpha1.Job {
+	t.Helper()
+	job := &executionv1alpha1.Job{}
+	if err := kubeClient.Get(ctx, key, job); err != nil {
+		t.Fatal(err)
+	}
+	return job
+}
+
+func conditionIs(job *executionv1alpha1.Job, conditionType string, status metav1.ConditionStatus) bool {
+	condition := meta.FindStatusCondition(job.Status.Conditions, conditionType)
+	return condition != nil && condition.Status == status
+}