feat(nvidia-tuned): update service to be eks to match aicr

Author: ayuskauskas

nvidia-tuned/README.md

@@ -8,7 +8,7 @@ This package inherits from the base `tuned` package and adds pre-configured tune
 
 - **Common base profiles**: Foundational settings deployed to `/usr/lib/tuned/`
 - **OS-specific workload profiles**: Profiles that may vary by OS version
-- **Service profiles**: Service-specific settings (AWS, GCP, etc.)
+- **Service profiles**: Service-specific settings (eks, GCP, etc.)
 
 The configmap uses an **intent-based** model where you specify **what** you want (intent + accelerator) rather than a specific profile name. The profile name `nvidia-{accelerator}-{intent}` is constructed automatically.
 
@@ -56,7 +56,7 @@ profiles/
 │   └── rhel/
 │       └── 9/              # Symlinks to os/common/ (override when needed)
 └── service/
-    └── aws/
+    └── eks/
         ├── tuned.conf.template  # Service template (include= added dynamically)
         └── script.sh
 ```
@@ -95,10 +95,10 @@ Examples:
 
 ### Inheritance Chain
 
-When you specify `intent: inference`, `accelerator: h100`, and `service: aws`:
+When you specify `intent: inference`, `accelerator: h100`, and `service: eks`:
 
 ```
-aws (active profile)
+eks (active profile)
   └── includes: nvidia-h100-inference
         └── includes: nvidia-h100-performance
               └── includes: nvidia-acs-disable
@@ -111,7 +111,7 @@ aws (active profile)
 apiVersion: skyhook.nvidia.com/v1alpha1
 kind: Skyhook
 metadata:
-  name: nvidia-tuned-aws
+  name: nvidia-tuned-eks
 spec:
   nodeSelectors:
     matchLabels:
@@ -131,7 +131,7 @@ spec:
       configMap:
         intent: inference
         accelerator: h100
-        service: aws
+        service: eks
 ```
 
 ### ConfigMap Fields
@@ -140,7 +140,7 @@ spec:
 |-------|----------|---------|-------------|
 | `accelerator` | Yes | — | GPU/accelerator type (e.g., `h100`) |
 | `intent` | No | `performance` | Workload intent (e.g., `inference`, `performance`, `multiNodeTraining`) |
-| `service` | No | — | Service name (e.g., `aws`). If specified, service profile wraps the workload profile |
+| `service` | No | — | Service name (e.g., `eks`). If specified, service profile wraps the workload profile |
 
 ## Available Profiles
 
@@ -163,7 +163,7 @@ spec:
 
 | Service | Description |
 |---------|-------------|
-| `aws` | AWS-specific settings (MAC address policy for CNI) |
+| `eks` | eks-specific settings (MAC address policy for CNI) |
 
 ## Adding OS-Specific Overrides
 

nvidia-tuned/config.json

@@ -1,7 +1,7 @@
 {
     "schema_version": "v1",
     "package_name": "nvidia_tuned",
-    "package_version": "0.2.0",
+    "package_version": "0.2.3",
     "expected_config_files": ["accelerator"],
     "modes": {
         "uninstall": [

…tuned/profiles/service/aws/bootloader.sh …tuned/profiles/service/eks/bootloader.shnvidia-tuned/profiles/service/aws/bootloader.sh renamed to nvidia-tuned/profiles/service/eks/bootloader.sh nvidia-tuned/profiles/service/aws/bootloader.sh renamed to nvidia-tuned/profiles/service/eks/bootloader.sh

@@ -14,7 +14,7 @@ EXPECTED_NETWORK_CONTENT='[Link]
 MACAddressPolicy=none
 '
 
-# Profile dir (script is in e.g. /etc/tuned/aws-{accelerator}-{intent}/)
+# Profile dir (script is in e.g. /etc/tuned/eks-{accelerator}-{intent}/)
 SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
 
 apply_network_dropin() {

…/service/aws/nvidia-gb200-inference.conf …/service/eks/nvidia-gb200-inference.confnvidia-tuned/profiles/service/aws/nvidia-gb200-inference.conf renamed to nvidia-tuned/profiles/service/eks/nvidia-gb200-inference.conf nvidia-tuned/profiles/service/aws/nvidia-gb200-inference.conf renamed to nvidia-tuned/profiles/service/eks/nvidia-gb200-inference.conf

@@ -7,7 +7,7 @@
 - prepare_nvidia_profiles does the right thing for all combinations of:
   - accelerator (h100, gb200)
   - intent (performance, inference, multiNodeTraining)
-  - service (aws, none)
+  - service (eks, none)
 - For AWS service, verifies grub config file is created correctly
 """
 
@@ -261,14 +261,14 @@ def test_prepare_nvidia_profiles_no_service(base_image, accelerator, intent):
 
 @pytest.mark.parametrize("accelerator", ["h100", "gb200"])
 @pytest.mark.parametrize("intent", ["performance", "inference", "multiNodeTraining"])
-def test_prepare_nvidia_profiles_with_aws_service(base_image, accelerator, intent):
-    """Test prepare_nvidia_profiles with AWS service for all combinations."""
+def test_prepare_nvidia_profiles_with_eks_service(base_image, accelerator, intent):
+    """Test prepare_nvidia_profiles with EKS service for all combinations."""
     runner = DockerTestRunner(package="nvidia-tuned", base_image=base_image)
     try:
         configmaps = {
             "accelerator": accelerator,
             "intent": intent,
-            "service": "aws",
+            "service": "eks",
         }
 
         # Create container by running script (this creates the container)
@@ -296,21 +296,21 @@ def test_prepare_nvidia_profiles_with_aws_service(base_image, accelerator, inten
 
         # Final profile name = {service}-{accelerator}-{intent}
         expected_workload_profile = f"nvidia-{accelerator}-{intent}"
-        expected_final_profile = f"aws-{accelerator}-{intent}"
-        assert_output_contains(result.stdout, "Requested service: aws")
+        expected_final_profile = f"eks-{accelerator}-{intent}"
+        assert_output_contains(result.stdout, "Requested service: eks")
         assert_output_contains(result.stdout, f"include={expected_workload_profile}")
         assert_output_contains(result.stdout, f"Final profile name: {expected_final_profile}")
 
-        # Verify service profile directory exists (final name = aws-{accelerator}-{intent})
+        # Verify service profile directory exists (final name = eks-{accelerator}-{intent})
         service_profile_exists = runner.file_exists(f"/etc/tuned/{expected_final_profile}/tuned.conf")
-        assert service_profile_exists, f"AWS service profile {expected_final_profile} was not deployed"
+        assert service_profile_exists, f"EKS service profile {expected_final_profile} was not deployed"
 
         # Verify service profile includes the workload profile
         service_profile_content = runner.get_file_contents(
             f"/etc/tuned/{expected_final_profile}/tuned.conf"
         )
         assert f"include={expected_workload_profile}" in service_profile_content, \
-            f"AWS profile does not include {expected_workload_profile}"
+            f"EKS profile does not include {expected_workload_profile}"
 
         # Verify tuned_profile file points to final profile ({service}-{accelerator}-{intent})
         tuned_profile_content = runner.get_file_contents(
@@ -319,28 +319,28 @@ def test_prepare_nvidia_profiles_with_aws_service(base_image, accelerator, inten
         assert tuned_profile_content.strip() == expected_final_profile, \
             f"tuned_profile should be '{expected_final_profile}', got: {tuned_profile_content!r}"
 
-        # For AWS, verify bootloader script exists in final profile dir
+        # For EKS, verify bootloader script exists in final profile dir
         bootloader_script_exists = runner.file_exists(
             f"/etc/tuned/{expected_final_profile}/bootloader.sh"
         )
-        assert bootloader_script_exists, "AWS bootloader.sh script was not deployed"
+        assert bootloader_script_exists, "EKS bootloader.sh script was not deployed"
 
         # Verify script.sh exists in final profile dir
         script_exists = runner.file_exists(f"/etc/tuned/{expected_final_profile}/script.sh")
-        assert script_exists, "AWS script.sh was not deployed"
+        assert script_exists, "EKS script.sh was not deployed"
 
     finally:
         runner.cleanup()
 
 
-def test_prepare_nvidia_profiles_aws_grub_config(base_image):
-    """Test that AWS service creates the correct grub config file."""
+def test_prepare_nvidia_profiles_eks_grub_config(base_image):
+    """Test that EKS service creates the correct grub config file."""
     runner = DockerTestRunner(package="nvidia-tuned", base_image=base_image)
     try:
         configmaps = {
             "accelerator": "h100",
             "intent": "inference",
-            "service": "aws",
+            "service": "eks",
         }
 
         # Create container directly
@@ -371,9 +371,9 @@ def test_prepare_nvidia_profiles_aws_grub_config(base_image):
         assert "TUNED_BOOT_CMDLINE=\"iommu=pt hugepages=8192\"" in bootcmdline_content, \
             "Bootcmdline file should contain the actual boot parameters (iommu=pt hugepages=8192)"
 
-        # Final profile name = aws-h100-inference for this test's configmaps
-        final_profile = "aws-h100-inference"
-        # Run the AWS bootloader script (skip update-grub if it fails)
+        # Final profile name = eks-h100-inference for this test's configmaps
+        final_profile = "eks-h100-inference"
+        # Run the EKS bootloader script (skip update-grub if it fails)
         bootloader_result = runner.container.exec_run(
             ["bash", "-c", f"/etc/tuned/{final_profile}/bootloader.sh || true"],
             workdir="/"
@@ -443,14 +443,14 @@ def test_prepare_nvidia_profiles_missing_accelerator(base_image):
         runner.cleanup()
 
 
-def test_prepare_nvidia_profiles_aws_service_specific_profile(base_image):
-    """Test that AWS service-specific inference profiles are used when available."""
+def test_prepare_nvidia_profiles_eks_service_specific_profile(base_image):
+    """Test that EKS service-specific inference profiles are used when available."""
     runner = DockerTestRunner(package="nvidia-tuned", base_image=base_image)
     try:
         configmaps = {
             "accelerator": "h100",
             "intent": "inference",
-            "service": "aws",
+            "service": "eks",
         }
 
         # Create container directly
@@ -464,29 +464,29 @@ def test_prepare_nvidia_profiles_aws_service_specific_profile(base_image):
 
         assert_exit_code(result, 0)
 
-        # Verify that AWS-specific inference profile was deployed
+        # Verify that EKS-specific inference profile was deployed
         # (it should overwrite the OS profile)
         inference_profile_content = runner.get_file_contents(
             "/etc/tuned/nvidia-h100-inference/tuned.conf"
         )
 
-        # AWS-specific profile should NOT have scheduler parameters set (they may be in comments)
+        # EKS-specific profile should NOT have scheduler parameters set (they may be in comments)
         # Check that they're not set as actual sysctl parameters (not commented out)
         import re
 
         # Check for uncommented kernel.sched_latency_ns= lines
         latency_pattern = r'^\s*kernel\.sched_latency_ns\s*='
         assert not re.search(latency_pattern, inference_profile_content, re.MULTILINE), \
-            "AWS-specific inference profile should not contain uncommented kernel.sched_latency_ns"
+            "EKS-specific inference profile should not contain uncommented kernel.sched_latency_ns"
 
         # Check for uncommented kernel.sched_min_granularity_ns= lines
         granularity_pattern = r'^\s*kernel\.sched_min_granularity_ns\s*='
         assert not re.search(granularity_pattern, inference_profile_content, re.MULTILINE), \
-            "AWS-specific inference profile should not contain uncommented kernel.sched_min_granularity_ns"
+            "EKS-specific inference profile should not contain uncommented kernel.sched_min_granularity_ns"
 
         # But should have vm.swappiness
         assert "vm.swappiness=1" in inference_profile_content, \
-            "AWS-specific inference profile should contain vm.swappiness=1"
+            "EKS-specific inference profile should contain vm.swappiness=1"
 
     finally:
         runner.cleanup()