fix: stop app if it is hami cause unschedule no wait (#2531)

* fix: stop app if it is hami cause unschedule

* ingore param from req if size=0

framework/app-service/.olares/config/cluster/deploy/appservice_deploy.yaml

@@ -170,7 +170,7 @@ spec:
       priorityClassName: "system-cluster-critical"
       containers:
       - name: app-service
-        image: beclab/app-service:0.5.1
+        image: beclab/app-service:0.5.2
         imagePullPolicy: IfNotPresent
         ports:
         - containerPort: 6755

framework/app-service/pkg/apiserver/handler_settings.go

@@ -729,6 +729,12 @@ func (h *Handler) getApplicationProviderList(req *restful.Request, resp *restful
 		return
 	}
 
+	// sys app does not have app config
+	if am.Spec.Config == "" {
+		resp.WriteAsJson([]providerRegistry{})
+		return
+	}
+
 	var appConfig appcfg.ApplicationConfig
 	err = am.GetAppConfig(&appConfig)
 	if err != nil {

framework/app-service/pkg/apiserver/handler_webhook.go

@@ -338,7 +338,10 @@ func (h *Handler) getGPUResourceTypeKey(gpuType string) string {
 	case utils.GB10ChipType:
 		return constants.NvidiaGB10GPU
 	case utils.AmdApuCardType:
-		return constants.AMDAPU
+		return constants.AMDGPU
+	case utils.AmdGpuCardType:
+		return constants.AMDGPU
+
 	default:
 		return ""
 	}

framework/app-service/pkg/appinstaller/helm_ops_install.go

@@ -31,6 +31,7 @@ import (
 	corev1 "k8s.io/api/core/v1"
 	apierrors "k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/fields"
 	"k8s.io/apimachinery/pkg/runtime/schema"
 	"k8s.io/apimachinery/pkg/types"
 	"k8s.io/client-go/dynamic"
@@ -587,7 +588,7 @@ func (h *HelmOps) isStartUp() (bool, error) {
 		}
 		klog.Infof("podSErvers: %v", podNames)
 
-		serverStarted, err := checkIfStartup(serverPods, true)
+		serverStarted, err := h.checkIfStartup(serverPods, true)
 		if err != nil {
 			klog.Errorf("v2 app %s server pods not ready: %v", h.app.AppName, err)
 			return false, err
@@ -606,7 +607,7 @@ func (h *HelmOps) isStartUp() (bool, error) {
 		return false, err
 	}
 
-	clientStarted, err := checkIfStartup(clientPods, false)
+	clientStarted, err := h.checkIfStartup(clientPods, false)
 	if err != nil {
 		return false, err
 	}
@@ -669,7 +670,7 @@ func (h *HelmOps) findServerPods() ([]corev1.Pod, error) {
 	return pods, nil
 }
 
-func checkIfStartup(pods []corev1.Pod, isServerSide bool) (bool, error) {
+func (h *HelmOps) checkIfStartup(pods []corev1.Pod, isServerSide bool) (bool, error) {
 	if len(pods) == 0 {
 		return false, errors.New("no pod found")
 	}
@@ -678,6 +679,16 @@ func checkIfStartup(pods []corev1.Pod, isServerSide bool) (bool, error) {
 	for _, pod := range pods {
 		creationTime := pod.GetCreationTimestamp()
 		pendingDuration := time.Since(creationTime.Time)
+		pendingKind, err := h.getPendingKind(&pod)
+		if err != nil {
+			return false, err
+		}
+		if pendingKind == "hami-scheduler" {
+			if isServerSide {
+				return false, errcode.ErrServerSidePodPending
+			}
+			return false, errcode.ErrPodPending
+		}
 
 		if pod.Status.Phase == corev1.PodPending && pendingDuration > time.Minute*10 {
 			if isServerSide {
@@ -703,6 +714,28 @@ func checkIfStartup(pods []corev1.Pod, isServerSide bool) (bool, error) {
 	return false, nil
 }
 
+func (h *HelmOps) getPendingKind(pod *corev1.Pod) (string, error) {
+	fieldSelector := fields.OneTermEqualSelector("involvedObject.name", pod.Name).String()
+	events, err := h.client.KubeClient.Kubernetes().CoreV1().Events(pod.Namespace).List(h.ctx, metav1.ListOptions{
+		FieldSelector: fieldSelector,
+	})
+	if err != nil {
+		return "", err
+	}
+	eventFrom := ""
+	for _, event := range events.Items {
+		if event.Reason == "FailedScheduling" {
+			if event.ReportingController != "" {
+				eventFrom = event.ReportingController
+			} else {
+				eventFrom = event.Source.Component
+			}
+			break
+		}
+	}
+	return eventFrom, nil
+}
+
 type applicationSettingsSubPolicy struct {
 	URI      string `json:"uri"`
 	Policy   string `json:"policy"`

framework/app-service/pkg/constants/constants.go

@@ -87,6 +87,7 @@ const (
 	NvidiaGPU     = "nvidia.com/gpu"
 	NvidiaGB10GPU = "nvidia.com/gb10"
 	AMDAPU        = "amd.com/apu"
+	AMDGPU        = "amd.com/gpu"
 
 	AuthorizationLevelOfPublic  = "public"
 	AuthorizationLevelOfPrivate = "private"

framework/app-service/pkg/images/client.go

@@ -160,7 +160,7 @@ func (imc *ImageManagerClient) PollDownloadProgress(ctx context.Context, am *app
 					if t == "" {
 						imageSize := maxImageSize
 						info := findImageSize(imageList, ref.Name)
-						if info != nil {
+						if info != nil && info.Size != 0 {
 							//klog.Infof("get image:%s size:%d", ref.Name, info.Size)
 							imageSize = info.Size
 						}