[CONTINT-4105] Support arbitrary container-ids to collect container metrics

.github/CODEOWNERS

@@ -409,6 +409,7 @@
 /pkg/util/pdhutil/                      @DataDog/windows-agent
 /pkg/util/winutil/                      @DataDog/windows-agent
 /pkg/util/testutil/flake                @DataDog/agent-developer-tools
+/pkg/util/trie                          @DataDog/container-integrations
 /pkg/languagedetection                  @DataDog/processes @DataDog/universal-service-monitoring
 /pkg/logs/                              @DataDog/agent-metrics-logs
 /pkg/logs/launchers/windowsevent        @DataDog/agent-metrics-logs @DataDog/windows-agent

comp/core/workloadmeta/collectors/internal/containerd/container_builder.go

@@ -135,6 +135,9 @@ func buildWorkloadMetaContainer(namespace string, container containerd.Container
 
 		workloadContainer.EnvVars = envs
 		workloadContainer.Hostname = spec.Hostname
+		if spec.Linux != nil {
+			workloadContainer.CgroupPath = extractCgroupPath(spec.Linux.CgroupsPath)
+		}
 	} else if errors.Is(err, cutil.ErrSpecTooLarge) {
 		log.Warnf("Skipping parsing of container spec for container id: %s, spec is bigger than: %d", info.ID, cutil.DefaultAllowedSpecMaxSize)
 	} else {
@@ -144,6 +147,17 @@ func buildWorkloadMetaContainer(namespace string, container containerd.Container
 	return workloadContainer, nil
 }
 
+// Containerd applies some transformations to the cgroup path, we need to revert them
+// https://github.com/containerd/containerd/blob/b168147ca8fccf05003117324f493d40f97b6077/internal/cri/server/podsandbox/helpers_linux.go#L64-L65
+// See https://github.com/opencontainers/runc/blob/main/docs/systemd.md
+func extractCgroupPath(path string) string {
+	res := path
+	if l := strings.Split(path, ":"); len(l) == 3 {
+		res = l[0] + "/" + l[1] + "-" + l[2] + ".scope"
+	}
+	return res
+}
+
 func extractStatus(status containerd.ProcessStatus) workloadmeta.ContainerStatus {
 	switch status {
 	case containerd.Paused, containerd.Pausing:

comp/core/workloadmeta/collectors/internal/containerd/container_builder_test.go

@@ -109,7 +109,9 @@ func TestBuildWorkloadMetaContainer(t *testing.T) {
 			}, nil
 		},
 		MockSpec: func(namespace string, ctn containers.Container) (*oci.Spec, error) {
-			return &oci.Spec{Hostname: hostName, Process: &specs.Process{Env: envVarStrs}}, nil
+			return &oci.Spec{Hostname: hostName, Process: &specs.Process{Env: envVarStrs}, Linux: &specs.Linux{
+				CgroupsPath: "kubelet-kubepods-burstable-pod99dcb84d2a34f7e338778606703258c4.slice:cri-containerd:ec9ea0ad54dd0d96142d5dbe11eb3f1509e12ba9af739620c7b5ad377ce94602",
+			}}, nil
 		},
 		MockStatus: func(namespace string, ctn containerd.Container) (containerd.ProcessStatus, error) {
 			return containerd.Running, nil
@@ -175,10 +177,49 @@ func TestBuildWorkloadMetaContainer(t *testing.T) {
 		NetworkIPs: make(map[string]string), // Not available
 		Hostname:   hostName,
 		PID:        0, // Not available
+		CgroupPath: "kubelet-kubepods-burstable-pod99dcb84d2a34f7e338778606703258c4.slice/cri-containerd-ec9ea0ad54dd0d96142d5dbe11eb3f1509e12ba9af739620c7b5ad377ce94602.scope",
 	}
 	assert.Equal(t, expected, result)
 }
 
+func TestExtractCgroupPath(t *testing.T) {
+	tests := []struct {
+		name     string
+		path     string
+		expected string
+	}{
+		// cgroupfs retrieved using minikube + qemu2 driver
+		{
+			name:     "cgroupfs + kubernetes pod",
+			path:     "/kubepods/burstable/pod84a7cac1-5690-4935-bffc-4b808e0240e4/0af39253daf5d1054519efdd054023e929785c2813c29f6a0ce887f652e1a997",
+			expected: "/kubepods/burstable/pod84a7cac1-5690-4935-bffc-4b808e0240e4/0af39253daf5d1054519efdd054023e929785c2813c29f6a0ce887f652e1a997",
+		},
+		// systemd retrieved using kind
+		{
+			name:     "systemd + kubernetes pod",
+			path:     "kubelet-kubepods-burstable-pod99dcb84d2a34f7e338778606703258c4.slice:cri-containerd:ec9ea0ad54dd0d96142d5dbe11eb3f1509e12ba9af739620c7b5ad377ce94602",
+			expected: "kubelet-kubepods-burstable-pod99dcb84d2a34f7e338778606703258c4.slice/cri-containerd-ec9ea0ad54dd0d96142d5dbe11eb3f1509e12ba9af739620c7b5ad377ce94602.scope",
+		},
+		// custom retrieved using ctr run
+		{
+			name:     "systemd/cgroupfs + container",
+			path:     "/default/redis",
+			expected: "/default/redis",
+		},
+		// garden
+		{
+			name:     "garden",
+			path:     "/garden/1a65c217-84b2-8d13-3d78-46b14b6c7ea1",
+			expected: "/garden/1a65c217-84b2-8d13-3d78-46b14b6c7ea1",
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			assert.Equal(t, tt.expected, extractCgroupPath(tt.path))
+		})
+	}
+}
+
 func TestExtractRuntimeFlavor(t *testing.T) {
 	tests := []struct {
 		name     string

comp/core/workloadmeta/dump_test.go

@@ -12,6 +12,7 @@ import (
 	"github.com/DataDog/datadog-agent/comp/core/config"
 	"github.com/DataDog/datadog-agent/comp/core/log/logimpl"
 	"github.com/DataDog/datadog-agent/pkg/util/fxutil"
+
 	"github.com/stretchr/testify/assert"
 	"go.uber.org/fx"
 )
@@ -59,7 +60,8 @@ func TestDump(t *testing.T) {
 		Image: ContainerImage{
 			Tag: "latest",
 		},
-		PID: 1,
+		PID:        1,
+		CgroupPath: "/default/ctr-id",
 	}
 
 	s.handleEvents([]CollectorEvent{
@@ -134,6 +136,7 @@ Allowed env variables: DD_SERVICE:my-svc DD_ENV:prod DD_VERSION:v1
 Hostname: 
 Network IPs: 
 PID: 0
+Cgroup path: 
 `,
 					"source:source2 id: ctr-id": `----------- Entity ID -----------
 Kind: container ID: ctr-id
@@ -163,6 +166,7 @@ Allowed env variables:
 Hostname: 
 Network IPs: 
 PID: 1
+Cgroup path: /default/ctr-id
 `,
 					"sources(merged):[source1 source2] id: ctr-id": `----------- Entity ID -----------
 Kind: container ID: ctr-id
@@ -192,6 +196,7 @@ Allowed env variables: DD_SERVICE:my-svc DD_ENV:prod DD_VERSION:v1
 Hostname: 
 Network IPs: 
 PID: 1
+Cgroup path: /default/ctr-id
 `,
 				},
 			},

comp/core/workloadmeta/types.go

@@ -522,6 +522,10 @@ type Container struct {
 	Owner           *EntityID
 	SecurityContext *ContainerSecurityContext
 	Resources       ContainerResources
+	// CgroupPath is a path to the cgroup of the container.
+	// It can be relative to the cgroup parent.
+	// Linux only.
+	CgroupPath string
 }
 
 // GetID implements Entity#GetID.
@@ -575,6 +579,7 @@ func (c Container) String(verbose bool) string {
 		_, _ = fmt.Fprintln(&sb, "Hostname:", c.Hostname)
 		_, _ = fmt.Fprintln(&sb, "Network IPs:", mapToString(c.NetworkIPs))
 		_, _ = fmt.Fprintln(&sb, "PID:", c.PID)
+		_, _ = fmt.Fprintln(&sb, "Cgroup path:", c.CgroupPath)
 	}
 
 	if len(c.Ports) > 0 && verbose {

comp/dogstatsd/server/convert_bench_test.go

@@ -78,5 +78,5 @@ type ServerDeps struct {
 }
 
 func newServerDeps(t testing.TB, options ...fx.Option) ServerDeps {
-	return fxutil.Test[ServerDeps](t, core.MockBundle(), workloadmeta.MockModule(), fx.Supply(workloadmeta.NewParams()), fx.Options(options...))
+	return fxutil.Test[ServerDeps](t, core.MockBundle(), workloadmeta.MockModuleV2(), fx.Supply(workloadmeta.NewParams()), fx.Options(options...))
 }

pkg/proto/datadog/workloadmeta/workloadmeta.proto

@@ -51,7 +51,7 @@ message ContainerImage {
   string name = 3;
   string shortName = 4;
   string tag = 5;
-  string repo_digest = 6; 
+  string repo_digest = 6;
 }
 
 message ContainerPort {
@@ -106,6 +106,7 @@ message Container {
   Runtime runtime = 9;
   ContainerState state = 10;
   repeated string collectorTags = 11;
+  string cgroupPath = 12;
 }
 
 message KubernetesPodOwner {