add DRAKWOKDriver dependency to install kwok and dra resources for simulated scale testing

clusterloader2/cmd/clusterloader.go

@@ -44,6 +44,7 @@ import (
 	"k8s.io/perf-tests/clusterloader2/pkg/util"
 
 	_ "k8s.io/perf-tests/clusterloader2/pkg/dependency/dra"
+	_ "k8s.io/perf-tests/clusterloader2/pkg/dependency/kwok/dra"
 	_ "k8s.io/perf-tests/clusterloader2/pkg/measurement/common"
 	_ "k8s.io/perf-tests/clusterloader2/pkg/measurement/common/bundle"
 	_ "k8s.io/perf-tests/clusterloader2/pkg/measurement/common/dns"

clusterloader2/pkg/dependency/kwok/README.md

@@ -0,0 +1,197 @@
+# KWOK DRA Dependency
+
+This dependency provides fake Kubernetes nodes with Dynamic Resource Allocation (DRA) GPU resources using [KWOK (Kubernetes WithOut Kubelet)](https://kwok.sigs.k8s.io/).
+
+## What it does
+
+- Installs KWOK controller in `kwok-system` namespace
+- Creates fake nodes with GPU resources exposed through DRA ResourceSlices
+- Enables testing DRA workloads without real GPU hardware
+
+## Configuration
+
+Add the dependency to your ClusterLoader2 test configuration:
+
+```yaml
+# In your CL2 test config
+dependencyConfigs:
+- name: kwok-dra
+  params:
+    nodes: 4              # Number of fake nodes (default: 2)
+    gpusPerNode: 16       # GPUs per node (default: 8)
+    timeout: "10m"        # Setup timeout (default: 5m)
+```
+
+## Fake Resources Created
+
+### Nodes
+- **Name**: `kwok-node-0`, `kwok-node-1`, etc.
+- **Resources**: 32 CPU, 256Gi memory, 110 pods
+- **Labels**: `type=kwok`, `kubernetes.io/hostname=kwok-node-N`
+- **Taints**: `kwok.x-k8s.io/node=fake:NoSchedule` (prevents real workloads)
+
+### GPU Resources (DRA)
+- **Driver**: `cl2-gpu.kwok.x-k8s.io`
+- **API Version**: `resource.k8s.io/v1beta2`
+- **ResourceSlices**: One per node with configurable GPU devices  
+- **Device Names**: `gpu0`, `gpu1`, `gpu2`, etc.
+- **Capacity**: Each device provides `1` GPU unit (`cl2-gpu.kwok.x-k8s.io/gpu`)
+- **Device Attributes**: 
+  - `gpu-type`: "kwok-gpu"
+  - `memory`: "8Gi" 
+  - `compute-capability`: "7.5"
+
+## Example: Simple GPU Job
+
+Create a job that requests fake GPUs using the provided example files:
+
+### 1. GPU Job Template
+See [`examples/kwok-gpu-job.yaml`](examples/kwok-gpu-job.yaml) - A ClusterLoader2 job template that:
+- Uses the `job-type: short-lived` labels for KWOK completion simulation
+- Includes proper tolerations for KWOK fake nodes
+- References the GPU ResourceClaimTemplate
+- Uses templating variables (`{{.Name}}`, `{{.Replicas}}`, etc.)
+
+### 2. ResourceClaimTemplate 
+See [`examples/kwok-gpu-resource-claim-template.yaml`](examples/kwok-gpu-resource-claim-template.yaml) - Defines:
+- `v1beta2` ResourceClaimTemplate for requesting GPU devices
+- References the `cl2-gpu.kwok.x-k8s.io` DeviceClass
+- Created first in a separate step before jobs are created
+
+## Running Tests with KWOK DRA
+
+### Using the Main E2E Script
+
+The `run-e2e.sh` script is the main entry point for running performance tests in the perf-tests repository.
+
+```bash
+# Basic usage from perf-tests root directory
+./run-e2e.sh <tool-name> [options...]
+
+# Run a ClusterLoader2 test with KWOK DRA dependency
+./run-e2e.sh cluster-loader2 \
+  --testconfig=pkg/dependency/kwok/examples/test-config.yaml \
+  --provider=skeleton \
+  --nodes=3 \
+  --report-dir=/tmp/reports
+
+# Quick test with different node counts
+./run-e2e.sh cluster-loader2 \
+  --testconfig=pkg/dependency/kwok/examples/test-config.yaml \
+  --provider=skeleton \
+  --nodes=5 \
+  --report-dir=/tmp/reports
+
+# View available test tools
+./run-e2e.sh --help
+```
+
+### Quick Start
+
+1. **Prerequisites**: Ensure you have a Kubernetes cluster running
+2. **Environment**: Set `KUBECONFIG` or `~/.kube/config` pointing to your cluster  
+3. **Run Test**: Execute the script with desired parameters
+4. **Results**: Check the `--report-dir` for test results and metrics
+
+### Available Test Tools
+
+The `run-e2e.sh` script supports multiple performance testing tools:
+
+- **`cluster-loader2`** - Kubernetes cluster performance and scale testing
+- **`network-performance`** - Network performance benchmarks  
+- **`kube-dns`** - DNS performance testing
+- **`core-dns`** - CoreDNS performance testing
+- **`node-local-dns`** - NodeLocalDNS performance testing
+
+### Example ClusterLoader2 Test Config
+
+Use the provided test configuration file:
+
+```bash
+# Copy the example config to your test directory
+cp pkg/dependency/kwok/examples/test-config.yaml your-test-config.yaml
+
+# Or reference it directly
+./run-e2e.sh cluster-loader2 \
+  --testconfig=pkg/dependency/kwok/examples/test-config.yaml \
+  --provider=kind \
+  --nodes=3 \
+  --report-dir=/tmp/kwok-dra-test
+```
+
+The [`examples/test-config.yaml`](examples/test-config.yaml) includes:
+- **KWOK DRA dependency** with 3 nodes and 8 GPUs per node  
+- **ResourceClaimTemplate creation** step (must run before jobs)
+- **10 GPU jobs** requesting fake GPU resources
+- **QPS throttling** for controlled job creation
+
+## Job Timing Configuration
+
+To control how long simulated jobs run before completing:
+
+```bash
+# Set job duration to 10 seconds (10000ms)
+export CL2_JOB_RUNNING_TIME_MS=10000
+
+# Run your ClusterLoader2 test
+./clusterloader2 --testconfig=test-config.yaml
+```
+
+This affects all jobs with `job-type: short-lived` labels running on KWOK nodes.
+
+## Important Notes
+
+1. **Tolerations Required**: Jobs must tolerate the `kwok.x-k8s.io/node=fake:NoSchedule` taint
+2. **DeviceClass**: Uses the built-in `cl2-gpu.kwok.x-k8s.io` DeviceClass provided by the dependency
+3. **Step Ordering**: ResourceClaimTemplate must be created before jobs that reference it
+4. **Resource Dependencies**: Jobs depend on both DeviceClass (from dependency) and ResourceClaimTemplate (from first step)
+5. **Job Labels**: Pods must have `job-type: short-lived` label for KWOK job completion simulation
+6. **Job Completion**: Set `CL2_JOB_RUNNING_TIME_MS` environment variable to control simulated job duration (default: 30000ms)
+7. **Device Attributes**: v1beta2 API provides rich device metadata (GPU type, memory, compute capability)  
+8. **Enhanced Scheduling**: ResourceSlices include proper labels for improved resource discovery
+9. **Fake Resources**: GPUs are simulated - no actual GPU operations occur  
+10. **Cleanup**: The dependency automatically cleans up when tests complete
+
+## Troubleshooting
+
+### Common Issues
+
+- **Nodes not ready**: Check KWOK controller logs in `kwok-system` namespace
+- **Jobs not scheduling**: Verify tolerations and DeviceClass configuration
+- **Timeout errors**: Increase the `timeout` parameter in dependency config
+- **ResourceClaimTemplate not found**: Ensure the ResourceClaimTemplate step runs before job creation
+
+### Testing the Setup
+
+```bash
+# Test KWOK DRA dependency with example config
+./run-e2e.sh cluster-loader2 \
+  --testconfig=clusterloader2/pkg/dependency/kwok/examples/test-config.yaml \
+  --provider=kind \
+  --nodes=3 \
+  --report-dir=/tmp/kwok-test
+
+# Check KWOK nodes are created
+kubectl get nodes -l type=kwok
+
+# Verify ResourceSlices are available  
+kubectl get resourceslices
+
+# Check DeviceClass is installed
+kubectl get deviceclasses cl2-gpu.kwok.x-k8s.io
+```
+
+### Debug Mode
+
+Enable verbose logging by setting environment variables:
+
+```bash
+export KLOG_V=2
+./run-e2e.sh cluster-loader2 --testconfig=... --v=2
+```
+
+## See Also
+
+- [KWOK Documentation](https://kwok.sigs.k8s.io/)
+- [Kubernetes DRA Documentation](https://kubernetes.io/docs/concepts/scheduling-eviction/dynamic-resource-allocation/)
+- [ClusterLoader2 Documentation](../../../docs/)