feat: enable CDI in containerd configuration (#6726)

Co-authored-by: Ganeshkumar Ashokavardhanan <[email protected]>
Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>

Author: 3 people

aks-node-controller/parser/helper_test.go

@@ -383,6 +383,7 @@ func Test_getContainerdConfig(t *testing.T) {
 oom_score = -999
 [plugins."io.containerd.grpc.v1.cri"]
   sandbox_image = ""
+  enable_cdi = true
   [plugins."io.containerd.grpc.v1.cri".containerd]
     default_runtime_name = "runc"
     [plugins."io.containerd.grpc.v1.cri".containerd.runtimes.runc]
@@ -415,6 +416,7 @@ oom_score = -999
 oom_score = -999
 [plugins."io.containerd.grpc.v1.cri"]
   sandbox_image = ""
+  enable_cdi = true
   [plugins."io.containerd.grpc.v1.cri".containerd]
     default_runtime_name = "nvidia-container-runtime"
     [plugins."io.containerd.grpc.v1.cri".containerd.runtimes.nvidia-container-runtime]
@@ -444,6 +446,7 @@ oom_score = -999
 oom_score = -999
 [plugins."io.containerd.grpc.v1.cri"]
   sandbox_image = ""
+  enable_cdi = true
   [plugins."io.containerd.grpc.v1.cri".containerd]
     default_runtime_name = "runc"
     [plugins."io.containerd.grpc.v1.cri".containerd.runtimes.runc]

aks-node-controller/parser/templates/containerd.toml.gtpl

@@ -3,6 +3,7 @@ oom_score = -999{{if getHasDataDir .KubeletConfig}}
 root = "{{.KubeletConfig.GetContainerDataDir}}"{{- end}}
 [plugins."io.containerd.grpc.v1.cri"]
   sandbox_image = "{{ .KubeBinaryConfig.GetPodInfraContainerImageUrl }}"
+  enable_cdi = true
   [plugins."io.containerd.grpc.v1.cri".containerd]
     {{- if .TeleportConfig.GetStatus }}
     snapshotter = "teleportd"

pkg/agent/baker.go

@@ -1431,6 +1431,7 @@ oom_score = -999{{if HasDataDir }}
 root = "{{GetDataDir}}"{{- end}}
 [plugins."io.containerd.grpc.v1.cri"]
   sandbox_image = "{{GetPodInfraContainerSpec}}"
+  enable_cdi = true
   [plugins."io.containerd.grpc.v1.cri".containerd]
     {{- if TeleportEnabled }}
     snapshotter = "teleportd"

pkg/agent/baker_test.go

@@ -1781,6 +1781,7 @@ testdomain567.com:53 {
 oom_score = -999
 [plugins."io.containerd.grpc.v1.cri"]
   sandbox_image = ""
+  enable_cdi = true
   [plugins."io.containerd.grpc.v1.cri".containerd]
     snapshotter = "overlaybd"
     disable_snapshot_annotations = false
@@ -1882,6 +1883,18 @@ oom_score = -999
 			}
 			config.ContainerService.Properties.AgentPoolProfiles[0].Distro = datamodel.AKSUbuntuContainerd2204
 		}, nil),
+		Entry("AKSUbuntu2204 with containerd and CDI enabled", "AKSUbuntu2204+Containerd+CDI", "1.24.2", func(config *datamodel.NodeBootstrappingConfiguration) {
+			config.ContainerService.Properties.AgentPoolProfiles[0].KubernetesConfig = &datamodel.KubernetesConfig{
+				ContainerRuntime: datamodel.Containerd,
+			}
+			config.ContainerService.Properties.AgentPoolProfiles[0].Distro = datamodel.AKSUbuntuContainerd2204
+			config.KubeletConfig = map[string]string{}
+		}, func(o *nodeBootstrappingOutput) {
+			Expect(o.vars["CONTAINERD_CONFIG_CONTENT"]).NotTo(BeEmpty())
+			containerdConfigFileContent, err := getBase64DecodedValue([]byte(o.vars["CONTAINERD_CONFIG_CONTENT"]))
+			Expect(err).To(BeNil())
+			Expect(containerdConfigFileContent).To(ContainSubstring("enable_cdi = true"))
+		}),
 		Entry("AKSUbuntu2204 containerd with multi-instance GPU", "AKSUbuntu2204+Containerd+MIG", "1.19.13",
 			func(config *datamodel.NodeBootstrappingConfiguration) {
 				config.ContainerService.Properties.AgentPoolProfiles[0].KubernetesConfig = &datamodel.KubernetesConfig{