feat(controller): Reconcile TriageRun jobs

api/v2/application_types.go

@@ -67,11 +67,89 @@ type ApplicationSpec struct {
 	Jobs                 []batchv1.Job                              `json:"jobs,omitempty"`
 	CronJobs             []batchv1.CronJob                          `json:"cronJobs,omitempty"`
 
+	// Triage declares the bounded diagnostic actions that may be requested for
+	// this application through TriageRun resources.
+	// +optional
+	Triage *ApplicationTriageSpec `json:"triage,omitempty"`
+
 	// HTTPRouteTemplate is the desired HTTPRoute spec. Nil means no HTTPRoute.
 	// +optional
 	HTTPRouteTemplate *HTTPRouteTemplateSpec `json:"httpRouteTemplate,omitempty"`
 }
 
+// ApplicationTriageSpec contains the shared diagnostic runner and the actions
+// exposed by an Application. TriageRun selects actions by name. The default
+// action runs the declared runner unchanged; named actions receive an explicit
+// --action selector.
+type ApplicationTriageSpec struct {
+	// ContainerName selects a container from the Application pod template. It
+	// may be omitted when the Application has exactly one container.
+	// +optional
+	ContainerName string `json:"containerName,omitempty"`
+
+	// Command replaces the selected container's entrypoint when non-empty.
+	// +kubebuilder:validation:MinItems=1
+	// +kubebuilder:validation:MaxItems=64
+	// +optional
+	Command []string `json:"command,omitempty"`
+
+	// Args replaces the selected container's arguments when non-empty. For a
+	// named action, the controller appends "--action", the selected action name,
+	// and then that action's Args. The default action only appends its Args.
+	// +kubebuilder:validation:MinItems=1
+	// +kubebuilder:validation:MaxItems=64
+	// +optional
+	Args []string `json:"args,omitempty"`
+
+	// Env adds or overrides environment variables inherited from the selected
+	// application container.
+	// +kubebuilder:validation:MaxItems=128
+	// +optional
+	Env []corev1.EnvVar `json:"env,omitempty"`
+
+	// Resources deliberately does not inherit the parent container's resource
+	// requirements. When omitted, the controller applies small bounded
+	// defaults suitable for diagnostics.
+	// +optional
+	Resources *corev1.ResourceRequirements `json:"resources,omitempty"`
+
+	// TimeoutSeconds is the Job execution deadline. Zero selects the controller
+	// default.
+	// +kubebuilder:validation:Minimum=1
+	// +kubebuilder:validation:Maximum=3600
+	// +optional
+	TimeoutSeconds int64 `json:"timeoutSeconds,omitempty"`
+
+	// Actions lists the stable action names and metadata exposed to callers.
+	// +kubebuilder:validation:MinItems=1
+	// +kubebuilder:validation:MaxItems=16
+	// +listType=map
+	// +listMapKey=name
+	Actions []TriageActionSpec `json:"actions"`
+}
+
+// TriageActionSpec describes one action exposed by the shared diagnostic
+// runner. Execution identity and resource settings remain on the parent
+// ApplicationTriageSpec so every action uses the same bounded runtime.
+type TriageActionSpec struct {
+	// Name is the stable identifier selected by TriageRun. "default" invokes the
+	// shared runner without an action selector; every other name is passed as
+	// "--action <name>".
+	Name TriageActionName `json:"name"`
+
+	// Description is human-readable help shown by clients such as Watchtower.
+	// +kubebuilder:validation:MaxLength=512
+	// +optional
+	Description string `json:"description,omitempty"`
+
+	// Args are appended after the optional action selector when starting the
+	// diagnostic runner. They are suitable for action-specific flags, not
+	// executable paths.
+	// +kubebuilder:validation:MaxItems=32
+	// +optional
+	Args []string `json:"args,omitempty"`
+}
+
 // HTTPRouteTemplateSpec contains the fields needed to build a Gateway API HTTPRoute.
 type HTTPRouteTemplateSpec struct {
 	ParentRefs []gatewayv1.ParentReference `json:"parentRefs"`

cmd/manager/main.go

@@ -40,6 +40,7 @@ import (
 	"github.com/wandb/operator/pkg/wandb/spec/channel/deployer"
 	corev1 "k8s.io/api/core/v1"
 	"k8s.io/client-go/discovery"
+	"k8s.io/client-go/kubernetes"
 	"sigs.k8s.io/controller-runtime/pkg/client/config"
 	gatewayv1 "sigs.k8s.io/gateway-api/apis/v1"
 
@@ -339,6 +340,22 @@ func main() {
 		os.Exit(1)
 	}
 
+	kubernetesClient, err := kubernetes.NewForConfig(mgr.GetConfig())
+	if err != nil {
+		setupLog.Error(err, "unable to create Kubernetes client", "controller", "TriageRun")
+		os.Exit(1)
+	}
+	if err = (&controller.TriageRunReconciler{
+		Client: mgr.GetClient(),
+		Scheme: mgr.GetScheme(),
+		PodLogs: &controller.KubernetesTriagePodLogReader{
+			CoreV1: kubernetesClient.CoreV1(),
+		},
+	}).SetupWithManager(mgr); err != nil {
+		setupLog.Error(err, "unable to create controller", "controller", "TriageRun")
+		os.Exit(1)
+	}
+
 	if enableWebhooks && enableV2 {
 		if err := webhookv2.SetupApplicationWebhookWithManager(mgr); err != nil {
 			setupLog.Error(err, "unable to create webhook", "webhook", "Application")

config/crd/bases/apps.wandb.com_applications.yaml

@@ -12801,6 +12801,170 @@ spec:
                   type:
                     type: string
                 type: object
+              triage:
+                properties:
+                  actions:
+                    items:
+                      properties:
+                        args:
+                          items:
+                            type: string
+                          maxItems: 32
+                          type: array
+                        description:
+                          maxLength: 512
+                          type: string
+                        name:
+                          minLength: 1
+                          type: string
+                      required:
+                      - name
+                      type: object
+                    maxItems: 16
+                    minItems: 1
+                    type: array
+                    x-kubernetes-list-map-keys:
+                    - name
+                    x-kubernetes-list-type: map
+                  args:
+                    items:
+                      type: string
+                    maxItems: 64
+                    minItems: 1
+                    type: array
+                  command:
+                    items:
+                      type: string
+                    maxItems: 64
+                    minItems: 1
+                    type: array
+                  containerName:
+                    type: string
+                  env:
+                    items:
+                      properties:
+                        name:
+                          type: string
+                        value:
+                          type: string
+                        valueFrom:
+                          properties:
+                            configMapKeyRef:
+                              properties:
+                                key:
+                                  type: string
+                                name:
+                                  default: ""
+                                  type: string
+                                optional:
+                                  type: boolean
+                              required:
+                              - key
+                              type: object
+                              x-kubernetes-map-type: atomic
+                            fieldRef:
+                              properties:
+                                apiVersion:
+                                  type: string
+                                fieldPath:
+                                  type: string
+                              required:
+                              - fieldPath
+                              type: object
+                              x-kubernetes-map-type: atomic
+                            fileKeyRef:
+                              properties:
+                                key:
+                                  type: string
+                                optional:
+                                  default: false
+                                  type: boolean
+                                path:
+                                  type: string
+                                volumeName:
+                                  type: string
+                              required:
+                              - key
+                              - path
+                              - volumeName
+                              type: object
+                              x-kubernetes-map-type: atomic
+                            resourceFieldRef:
+                              properties:
+                                containerName:
+                                  type: string
+                                divisor:
+                                  anyOf:
+                                  - type: integer
+                                  - type: string
+                                  pattern: ^(\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))))?$
+                                  x-kubernetes-int-or-string: true
+                                resource:
+                                  type: string
+                              required:
+                              - resource
+                              type: object
+                              x-kubernetes-map-type: atomic
+                            secretKeyRef:
+                              properties:
+                                key:
+                                  type: string
+                                name:
+                                  default: ""
+                                  type: string
+                                optional:
+                                  type: boolean
+                              required:
+                              - key
+                              type: object
+                              x-kubernetes-map-type: atomic
+                          type: object
+                      required:
+                      - name
+                      type: object
+                    maxItems: 128
+                    type: array
+                  resources:
+                    properties:
+                      claims:
+                        items:
+                          properties:
+                            name:
+                              type: string
+                            request:
+                              type: string
+                          required:
+                          - name
+                          type: object
+                        type: array
+                        x-kubernetes-list-map-keys:
+                        - name
+                        x-kubernetes-list-type: map
+                      limits:
+                        additionalProperties:
+                          anyOf:
+                          - type: integer
+                          - type: string
+                          pattern: ^(\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))))?$
+                          x-kubernetes-int-or-string: true
+                        type: object
+                      requests:
+                        additionalProperties:
+                          anyOf:
+                          - type: integer
+                          - type: string
+                          pattern: ^(\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))(([KMGTPE]i)|[numkMGTPE]|([eE](\+|-)?(([0-9]+(\.[0-9]*)?)|(\.[0-9]+))))?$
+                          x-kubernetes-int-or-string: true
+                        type: object
+                    type: object
+                  timeoutSeconds:
+                    format: int64
+                    maximum: 3600
+                    minimum: 1
+                    type: integer
+                required:
+                - actions
+                type: object
               volumeClaimTemplates:
                 items:
                   properties: