Changed shared GPU configmap name suffix from runai-sh-gpu to shared-gpu (#283)

* Changed shared GPU configmap name suffix from runai-sh-gpu to shared-gpu

* Changed sharedGPUConfigMapNamePrefix to sharedGPUConfigMapAnnotation

Author: romanbaron

cmd/podgroupcontroller/app/app.go

@@ -65,7 +65,7 @@ func Run() error {
 	cacheOptions := cache.Options{}
 	cacheOptions.ByObject = map[client.Object]cache.ByObject{
 		&v1.Pod{}:                     {Field: schedulerSelector},
-		&v1.Node{}:                    {}, // TODO: filter by strict/non-strict runai nodes
+		&v1.Node{}:                    {},
 		&schedulingv1.PriorityClass{}: {},
 		&v2alpha2.PodGroup{}:          {},
 	}

pkg/binder/binding/fraction_binder_test.go

@@ -46,7 +46,7 @@ var happyFlowObjectsBc = []runtime.Object{
 						ValueFrom: &v1.EnvVarSource{
 							ConfigMapKeyRef: &v1.ConfigMapKeySelector{
 								LocalObjectReference: v1.LocalObjectReference{
-									Name: "my-configmap-runai-sh-gpu",
+									Name: "my-configmap-shared-gpu",
 								},
 							},
 						},
@@ -58,7 +58,7 @@ var happyFlowObjectsBc = []runtime.Object{
 					VolumeSource: v1.VolumeSource{
 						ConfigMap: &v1.ConfigMapVolumeSource{
 							LocalObjectReference: v1.LocalObjectReference{
-								Name: "my-configmap-runai-sh-gpu",
+								Name: "my-configmap-shared-gpu",
 							},
 						},
 					},
@@ -102,7 +102,7 @@ var happyFlowObjects = []runtime.Object{
 					VolumeSource: v1.VolumeSource{
 						ConfigMap: &v1.ConfigMapVolumeSource{
 							LocalObjectReference: v1.LocalObjectReference{
-								Name: "my-configmap-runai-sh-gpu",
+								Name: "my-configmap-shared-gpu",
 							},
 						},
 					},
@@ -209,7 +209,7 @@ var _ = Describe("FractionBinder", func() {
 
 				configMap := &v1.ConfigMap{
 					ObjectMeta: metav1.ObjectMeta{
-						Name:      "my-configmap-runai-sh-gpu",
+						Name:      "my-configmap-shared-gpu",
 						Namespace: "my-ns",
 					},
 				}

pkg/binder/common/gpusharingconfigmap/config_map.go

@@ -7,7 +7,6 @@ import (
 	"context"
 	"fmt"
 	"strconv"
-	"strings"
 
 	v1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/api/errors"
@@ -18,11 +17,11 @@ import (
 )
 
 const (
-	GPUSharingConfigMap         = "runai-sh-gpu"
-	DesiredConfigMapPrefixKey   = "runai/shared-gpu-configmap"
-	maxVolumeNameLength         = 63
-	configMapNameNumRandomChars = 7
-	configMapNameExtraChars     = configMapNameNumRandomChars + 6
+	gpuSharingConfigMapAnnotation = "runai/shared-gpu-configmap"
+	gpuSharingConfigMap           = "shared-gpu"
+	maxVolumeNameLength           = 63
+	configMapNameNumRandomChars   = 7
+	configMapNameExtraChars       = configMapNameNumRandomChars + 6
 )
 
 func UpsertJobConfigMap(ctx context.Context,
@@ -107,7 +106,7 @@ func patchConfigMap(
 }
 
 func SetGpuCapabilitiesConfigMapName(pod *v1.Pod, containerIndex int, containerType ContainerType) string {
-	namePrefix, found := pod.Annotations[DesiredConfigMapPrefixKey]
+	namePrefix, found := pod.Annotations[gpuSharingConfigMapAnnotation]
 	if !found {
 		namePrefix = generateConfigMapNamePrefix(pod, containerIndex)
 		setConfigMapNameAnnotation(pod, namePrefix)
@@ -128,7 +127,7 @@ func generateConfigMapNamePrefix(pod *v1.Pod, containerIndex int) string {
 	}
 	// volume name is the `${configMapName}-vol` and should be up to 63 bytes long,
 	// 4 for "-vol" , 7 random chars, and 2 hyphens - 13 in total
-	maxBaseNameLength := (maxVolumeNameLength - configMapNameExtraChars) - len(GPUSharingConfigMap)
+	maxBaseNameLength := (maxVolumeNameLength - configMapNameExtraChars) - len(gpuSharingConfigMap)
 	// also remove from the max length for "-{containerIndex}" or "-i{initContainerIndex}" in the name
 	maxBaseNameLength = maxBaseNameLength - len(strconv.Itoa(containerIndex)) - 1
 	// also allow for appending "-evar" in case of envFrom config map
@@ -137,7 +136,7 @@ func generateConfigMapNamePrefix(pod *v1.Pod, containerIndex int) string {
 		baseName = baseName[:maxBaseNameLength]
 	}
 	return fmt.Sprintf("%v-%v-%v", baseName,
-		utilrand.String(configMapNameNumRandomChars), GPUSharingConfigMap)
+		utilrand.String(configMapNameNumRandomChars), gpuSharingConfigMap)
 }
 
 func ExtractCapabilitiesConfigMapName(pod *v1.Pod, containerIndex int, containerType ContainerType) (string, error) {
@@ -146,7 +145,7 @@ func ExtractCapabilitiesConfigMapName(pod *v1.Pod, containerIndex int, container
 		containerIndexStr = "i" + containerIndexStr
 	}
 
-	namePrefix, found := pod.Annotations[DesiredConfigMapPrefixKey]
+	namePrefix, found := pod.Annotations[gpuSharingConfigMapAnnotation]
 	if !found {
 		return "", fmt.Errorf("no desired configmap name found in pod %s/%s annotations", pod.Namespace, pod.Name)
 	}
@@ -162,59 +161,11 @@ func ExtractDirectEnvVarsConfigMapName(pod *v1.Pod, containerIndex int, containe
 	return fmt.Sprintf("%s-evar", configNameBase), nil
 }
 
-func HandleBCPod(ctx context.Context, kubeClient client.Client, pod *v1.Pod) (string, error) {
-	logger := log.FromContext(ctx)
-	desiredConfigMapName, err := GetDesiredConfigMapNameBC(pod)
-	if err != nil {
-		return "", err
-	}
-	logger.Info("Desired configmap name for backwards compatibility pod",
-		"namespace", pod.Namespace, "name", pod.Name, "configMapName", desiredConfigMapName)
-	err = UpdateBCPod(ctx, kubeClient, pod, desiredConfigMapName)
-	return desiredConfigMapName, err
-}
-
-func GetDesiredConfigMapNameBC(pod *v1.Pod) (string, error) {
-	cmName := ""
-	for _, volume := range pod.Spec.Volumes {
-		if volume.ConfigMap == nil {
-			continue
-		}
-
-		possibleCmName := volume.ConfigMap.LocalObjectReference.Name
-		if strings.HasSuffix(possibleCmName, GPUSharingConfigMap) {
-			if cmName != "" {
-				return "", fmt.Errorf("multiple desired gpu sharing configmap volumes detected for backwards "+
-					"compatibility pod %s/%s", pod.Namespace, pod.Name)
-			}
-			cmName = possibleCmName
-		}
-	}
-
-	if cmName == "" {
-		return "", fmt.Errorf("no desired gpu sharing configmap volume detected for backwards compatibility pod "+
-			"%s/%s", pod.Namespace, pod.Name)
-	}
-
-	return cmName, nil
-}
-
-func UpdateBCPod(ctx context.Context, kubeClient client.Client, pod *v1.Pod, desiredConfigMapName string) error {
-	updatedPod := pod.DeepCopy()
-	setConfigMapNameAnnotation(updatedPod, desiredConfigMapName)
-	err := kubeClient.Patch(ctx, updatedPod, client.MergeFrom(pod))
-	if err != nil {
-		return fmt.Errorf("failed to update pod %v/%v with desired configmap name %v, error: %v",
-			pod.Namespace, pod.Name, desiredConfigMapName, err)
-	}
-	return nil
-}
-
 func setConfigMapNameAnnotation(pod *v1.Pod, name string) {
 	if pod.Annotations == nil {
 		pod.Annotations = map[string]string{}
 	}
-	pod.Annotations[DesiredConfigMapPrefixKey] = name
+	pod.Annotations[gpuSharingConfigMapAnnotation] = name
 }
 
 // ownerReferencesDifferent compares two OwnerReferences and returns true if they are not the same