add support for DRAExtendedResources

Signed-off-by: Alay Patel <alayp@nvidia.com>

Author: alaypatel07

clusterloader2/pkg/dependency/dra/dra.go

@@ -70,6 +70,10 @@ func (d *draDependency) Setup(config *dependency.Config) error {
 		"Namespace":       namespace,
 		"WorkerNodeCount": getWorkerCount(config),
 	}
+
+	if extendedResourceName, ok := config.Params["ExtendedResourceName"]; ok {
+		mapping["ExtendedResourceName"] = extendedResourceName
+	}
 	if err := config.ClusterFramework.ApplyTemplatedManifests(
 		manifestsFS,
 		"manifests/*.yaml",

clusterloader2/pkg/dependency/dra/manifests/deviceclass.yaml

@@ -1,10 +1,13 @@
 ---
 # Source: dra-example-driver/templates/deviceclass.yaml
-apiVersion: resource.k8s.io/v1beta1
+apiVersion: resource.k8s.io/v1
 kind: DeviceClass
 metadata:
   name: gpu.example.com
 spec:
   selectors:
   - cel: 
       expression: "device.driver == 'gpu.example.com'"
+  {{- if .ExtendedResourceName}}
+  extendedResourceName: "{{.ExtendedResourceName}}"
+  {{- end}}

clusterloader2/testing/dra/config.yaml

@@ -23,8 +23,15 @@
 {{$smallJobSize := 1}}
 {{$smallJobCompletions := 10}}
 {{$jobRunningTime := DefaultParam .CL2_JOB_RUNNING_TIME "30s"}}
+{{$ENABLE_EXTENDED_RESOURCES := DefaultParam .CL2_ENABLE_EXTENDED_RESOURCES false}}
 
+{{$extendedResourceName := ""}}
+{{if $ENABLE_EXTENDED_RESOURCES}}
+{{$extendedResourceName = DefaultParam .CL2_EXTENDED_RESOURCE_NAME "example.com/gpu"}}
+name: dra-extended-resources-steady-state
+{{else}}
 name: dra-steady-state
+{{end}}
 
 namespace:
   number: {{$namespaces}}
@@ -37,11 +44,15 @@ tuningSets:
   qpsLoad:
     qps: {{$STEADY_STATE_QPS}}
 
+
 dependencies:
 - name: Install dra-example-driver for test
   Method: DRATestDriver
   Params:
     WorkerNodeCount: {{.Nodes}}
+    {{if $ENABLE_EXTENDED_RESOURCES}}
+    ExtendedResourceName: {{$extendedResourceName}}
+    {{end}}
   Timeout: 5m
 
 steps:
@@ -89,6 +100,7 @@ steps:
             query: histogram_quantile(0.99, sum(rate(dra_grpc_operations_duration_seconds_bucket{method_name=~".*NodePrepareResources"}[%v])) by (le))
           - name: p99_dra_grpc_node_unprepare_resources
             query: histogram_quantile(0.99, sum(rate(dra_grpc_operations_duration_seconds_bucket{method_name=~".*NodeUnprepareResources"}[%v])) by (le))
+{{if not $ENABLE_EXTENDED_RESOURCES}}
 - name: Create ResourceClaimTemplates in namespaces
   phases:
     - namespaceRange:
@@ -99,6 +111,7 @@ steps:
       objectBundle:
       - basename: single-gpu
         objectTemplatePath: "resourceclaimtemplate.yaml"
+{{end}}
 - name: Fill cluster to {{$fillPercentage}}% utilization
   phases:
     - namespaceRange:
@@ -113,6 +126,11 @@ steps:
           Replicas: {{$longJobSize}}
           Mode: {{$MODE}}
           Sleep: {{$longJobRunningTime}}
+          {{if $ENABLE_EXTENDED_RESOURCES}}
+          ExtendedResource: {{ true }}
+          {{else}}
+          ExtendedResource: {{ false }}
+          {{end}}
 - name: Wait for fill pods to be running
   measurements:
     - Identifier: WaitForControlledPodsRunning
@@ -188,6 +206,11 @@ steps:
           CompletionReplicas: {{$smallJobCompletions}}
           Mode: {{$MODE}}
           Sleep: {{$jobRunningTime}}
+          {{if $ENABLE_EXTENDED_RESOURCES}}
+          ExtendedResource: {{ true }}
+          {{else}}
+          ExtendedResource: {{ false }} 
+          {{end}}
 - name: Wait for short-lived jobs to finish
   measurements:
     - Identifier: WaitForFinishedJobs

clusterloader2/testing/dra/job.yaml

@@ -23,8 +23,15 @@ spec:
           args:
             - {{.Sleep}}
           resources:
+          {{ if not .ExtendedResource }}
             claims:
             - name: gpu
+          {{ else }}
+            limits:
+              example.com/gpu: "1"
+          {{ end }}
+      {{ if not .ExtendedResource }}
       resourceClaims:
       - name: gpu
-        resourceClaimTemplateName: single-gpu-0
+        resourceClaimTemplateName: single-gpu-0
+      {{ end }}

clusterloader2/testing/dra/long-running-job.yaml

@@ -23,8 +23,15 @@ spec:
           args:
             - {{.Sleep}}
           resources:
+          {{ if not .ExtendedResource }}
             claims:
             - name: gpu
+          {{ else }}
+            limits:
+              example.com/gpu: "1"
+          {{ end }}
+      {{ if not .ExtendedResource }}
       resourceClaims:
       - name: gpu
-        resourceClaimTemplateName: single-gpu-0 
+        resourceClaimTemplateName: single-gpu-0
+      {{ end }}