address PR comments and keep things similar to upstream

Signed-off-by: Swati Gupta <[email protected]>

Author: guptaNswati

cmd/compute-domain-kubelet-plugin/driver.go

@@ -139,6 +139,7 @@ func (d *driver) Shutdown() error {
 	if d == nil {
 		return nil
 	}
+
 	if err := d.state.computeDomainManager.Stop(); err != nil {
 		return fmt.Errorf("error stopping ComputeDomainManager: %w", err)
 	}

cmd/compute-domain-kubelet-plugin/health.go

@@ -1,4 +1,5 @@
 /*
+ * Copyright 2025 The Kubernetes Authors.
  * Copyright (c) 2025, NVIDIA CORPORATION.  All rights reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
@@ -61,7 +62,7 @@ func startHealthcheck(ctx context.Context, config *Config) (*healthcheck, error)
 		Scheme: "unix",
 		// TODO: this needs to adapt when seamless upgrades
 		// are enabled and the filename includes a uid.
-		Path: path.Join(DriverRegistrarPath, DriverName+"-reg.sock"),
+		Path: path.Join(config.flags.kubeletRegistrarDirectoryPath, DriverName+"-reg.sock"),
 	}).String()
 	klog.V(6).Infof("connecting to registration socket path=%s", regSockPath)
 	regConn, err := grpc.NewClient(
@@ -74,7 +75,7 @@ func startHealthcheck(ctx context.Context, config *Config) (*healthcheck, error)
 
 	draSockPath := (&url.URL{
 		Scheme: "unix",
-		Path:   path.Join(DriverPluginPath, "dra.sock"),
+		Path:   path.Join(config.DriverPluginPath(), "dra.sock"),
 	}).String()
 	klog.V(6).Infof("connecting to DRA socket path=%s", draSockPath)
 	draConn, err := grpc.NewClient(
@@ -106,8 +107,8 @@ func startHealthcheck(ctx context.Context, config *Config) (*healthcheck, error)
 }
 
 func (h *healthcheck) Stop() {
-	klog.Info("Stopping healthcheck service")
 	if h.server != nil {
+		klog.Info("Stopping healthcheck service")
 		h.server.GracefulStop()
 	}
 	h.wg.Wait()

cmd/compute-domain-kubelet-plugin/main.go

@@ -36,7 +36,6 @@ import (
 const (
 	DriverName                         = "compute-domain.nvidia.com"
 	DriverPluginCheckpointFileBasename = "checkpoint.json"
-	DriverRegistrarPath                = "/var/lib/kubelet/plugins_registry"
 )
 
 type Flags struct {

cmd/gpu-kubelet-plugin/driver.go

@@ -104,6 +104,7 @@ func (d *driver) Shutdown() error {
 	if d == nil {
 		return nil
 	}
+
 	if d.healthcheck != nil {
 		d.healthcheck.Stop()
 	}

cmd/gpu-kubelet-plugin/health.go

@@ -1,4 +1,5 @@
 /*
+ * Copyright 2025 The Kubernetes Authors.
  * Copyright (c) 2025, NVIDIA CORPORATION.  All rights reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
@@ -61,7 +62,7 @@ func startHealthcheck(ctx context.Context, config *Config) (*healthcheck, error)
 		Scheme: "unix",
 		// TODO: this needs to adapt when seamless upgrades
 		// are enabled and the filename includes a uid.
-		Path: path.Join(DriverRegistrarPath, DriverName+"-reg.sock"),
+		Path: path.Join(config.flags.kubeletRegistrarDirectoryPath, DriverName+"-reg.sock"),
 	}).String()
 	klog.V(6).Infof("connecting to registration socket path=%s", regSockPath)
 	regConn, err := grpc.NewClient(
@@ -74,7 +75,7 @@ func startHealthcheck(ctx context.Context, config *Config) (*healthcheck, error)
 
 	draSockPath := (&url.URL{
 		Scheme: "unix",
-		Path:   path.Join(DriverPluginPath, "dra.sock"),
+		Path:   path.Join(config.DriverPluginPath(), "dra.sock"),
 	}).String()
 	klog.V(6).Infof("connecting to DRA socket path=%s", draSockPath)
 	draConn, err := grpc.NewClient(
@@ -106,8 +107,8 @@ func startHealthcheck(ctx context.Context, config *Config) (*healthcheck, error)
 }
 
 func (h *healthcheck) Stop() {
-	klog.Info("Stopping healthcheck service")
 	if h.server != nil {
+		klog.Info("Stopping healthcheck service")
 		h.server.GracefulStop()
 	}
 	h.wg.Wait()

cmd/gpu-kubelet-plugin/main.go

@@ -36,7 +36,6 @@ import (
 const (
 	DriverName                         = "gpu.nvidia.com"
 	DriverPluginCheckpointFileBasename = "checkpoint.json"
-	DriverRegistrarPath                = "/var/lib/kubelet/plugins_registry"
 )
 
 type Flags struct {

deployments/helm/nvidia-dra-driver-gpu/values.yaml

@@ -154,14 +154,14 @@ kubeletPlugin:
       resources: {}
       # Port running a gRPC health service checked by a livenessProbe.
       # Set to a negative value to disable the service and the probe.
-      healthcheckPort: 61619
+      healthcheckPort: 51515
     gpus:
       securityContext:
         privileged: true
       resources: {}
       # Port running a gRPC health service checked by a livenessProbe.
       # Set to a negative value to disable the service and the probe.
-      healthcheckPort: 51518
+      healthcheckPort: 51516
   affinity:
     nodeAffinity:
       requiredDuringSchedulingIgnoredDuringExecution: