CD plugin: always inject CD details via CDI

Signed-off-by: Dr. Jan-Philip Gehrcke <[email protected]>

Author: jgehrcke

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

@@ -59,7 +59,8 @@ type ComputeDomainManager struct {
 }
 
 type ComputeDomainDaemonSettings struct {
-	manager         *ComputeDomainManager
+	manager *ComputeDomainManager
+	// CD UUID
 	domain          string
 	rootDir         string
 	configTmplPath  string
@@ -159,22 +160,8 @@ func (s *ComputeDomainDaemonSettings) GetDomain() string {
 }
 
 func (s *ComputeDomainDaemonSettings) GetCDIContainerEdits(ctx context.Context, devRoot string, info *nvcapDeviceInfo) (*cdiapi.ContainerEdits, error) {
-	cd, err := s.manager.GetComputeDomain(ctx, s.domain)
-	if err != nil {
-		return nil, fmt.Errorf("error getting compute domain: %w", err)
-	}
-	if cd == nil {
-		return nil, fmt.Errorf("compute domain not found: %s", s.domain)
-	}
-
 	edits := &cdiapi.ContainerEdits{
 		ContainerEdits: &cdispec.ContainerEdits{
-			Env: []string{
-				fmt.Sprintf("CLIQUE_ID=%s", s.manager.cliqueID),
-				fmt.Sprintf("COMPUTE_DOMAIN_UUID=%s", cd.UID),
-				fmt.Sprintf("COMPUTE_DOMAIN_NAME=%s", cd.Name),
-				fmt.Sprintf("COMPUTE_DOMAIN_NAMESPACE=%s", cd.Namespace),
-			},
 			Mounts: []*cdispec.Mount{
 				{
 					ContainerPath: "/etc/nvidia-imex",
@@ -193,7 +180,6 @@ func (s *ComputeDomainDaemonSettings) GetCDIContainerEdits(ctx context.Context,
 			},
 		},
 	}
-
 	return edits, nil
 }
 

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

@@ -30,6 +30,7 @@ import (
 	"k8s.io/klog/v2"
 	"k8s.io/kubernetes/pkg/kubelet/checkpointmanager"
 	cdiapi "tags.cncf.io/container-device-interface/pkg/cdi"
+	cdispec "tags.cncf.io/container-device-interface/specs-go"
 
 	configapi "github.com/NVIDIA/k8s-dra-driver-gpu/api/nvidia.com/resource/v1beta1"
 	"github.com/NVIDIA/k8s-dra-driver-gpu/pkg/featuregates"
@@ -491,6 +492,29 @@ func (s *DeviceState) applyComputeDomainDaemonConfig(ctx context.Context, config
 		ComputeDomain: config.DomainID,
 	}
 
+	cd, err := s.computeDomainManager.GetComputeDomain(ctx, config.DomainID)
+	if err != nil {
+		return nil, fmt.Errorf("error getting compute domain: %w", err)
+	}
+	if cd == nil {
+		return nil, fmt.Errorf("compute domain not found: %s", config.DomainID)
+	}
+
+	// Always inject Compute Domain details into the CD daemon, in a
+	// heterogeneous CD clique ID below may be empty (while other CD details are
+	// set, and consumed by the CD daemon).
+	edits := &cdiapi.ContainerEdits{
+		ContainerEdits: &cdispec.ContainerEdits{
+			Env: []string{
+				fmt.Sprintf("CLIQUE_ID=%s", s.computeDomainManager.cliqueID),
+				fmt.Sprintf("COMPUTE_DOMAIN_UUID=%s", cd.UID),
+				fmt.Sprintf("COMPUTE_DOMAIN_NAME=%s", cd.Name),
+				fmt.Sprintf("COMPUTE_DOMAIN_NAMESPACE=%s", cd.Namespace),
+			},
+		},
+	}
+	configState.containerEdits = configState.containerEdits.Append(edits)
+
 	// Only prepare files to inject to the daemon if IMEX is supported.
 	if s.computeDomainManager.cliqueID != "" {
 		// Parse the device node info for the fabic-imex-mgmt nvcap.