metrics: introduce client config to include alloc metadata as part of the base labels (#23964)

.changelog/23964.txt

@@ -0,0 +1,3 @@
+```release-note:improvement
+metrics: introduce client config to include alloc metadata as part of the base labels
+```

client/allocrunner/taskrunner/task_runner.go

@@ -510,6 +510,20 @@ func (tr *TaskRunner) initLabels() {
 		},
 	}
 
+	if tr.clientConfig.IncludeAllocMetadataInMetrics {
+		combined := alloc.Job.CombinedTaskMeta(alloc.TaskGroup, tr.taskName)
+		for meta, metaValue := range combined {
+			if len(tr.clientConfig.AllowedMetadataKeysInMetrics) > 0 && !slices.Contains(tr.clientConfig.AllowedMetadataKeysInMetrics, meta) {
+				continue
+			}
+
+			tr.baseLabels = append(tr.baseLabels, metrics.Label{
+				Name:  strings.ReplaceAll(meta, "-", "_"),
+				Value: metaValue,
+			})
+		}
+	}
+
 	if tr.alloc.Job.ParentID != "" {
 		tr.baseLabels = append(tr.baseLabels, metrics.Label{
 			Name:  "parent_id",

client/allocrunner/taskrunner/task_runner_test.go

@@ -2866,6 +2866,40 @@ func TestTaskRunner_BaseLabels(t *testing.T) {
 	require.Equal(alloc.Namespace, labels["namespace"])
 }
 
+// TestTaskRunner_BaseLabels_IncludesAllocMetadata tests that the base labels include
+// the allocation metadata fields using the provided allowed list of keys
+func TestTaskRunner_BaseLabels_IncludesAllocMetadata(t *testing.T) {
+	ci.Parallel(t)
+
+	alloc := mock.BatchAlloc()
+	alloc.Namespace = "not-default"
+	job := alloc.Job
+	job.Meta = map[string]string{"owner": "HashiCorp", "my-key": "my-value", "some_dynamic_value": "now()"}
+	task := job.TaskGroups[0].Tasks[0]
+	task.Driver = "raw_exec"
+	task.Config = map[string]interface{}{
+		"command": "whoami",
+	}
+
+	trConfig, cleanup := testTaskRunnerConfig(t, alloc, task.Name, nil)
+	defer cleanup()
+
+	trConfig.ClientConfig.IncludeAllocMetadataInMetrics = true
+	trConfig.ClientConfig.AllowedMetadataKeysInMetrics = []string{"owner", "my-key"}
+
+	tr, err := NewTaskRunner(trConfig)
+	must.NoError(t, err)
+
+	labels := map[string]string{}
+	for _, e := range tr.baseLabels {
+		labels[e.Name] = e.Value
+	}
+
+	must.Eq(t, "HashiCorp", labels["owner"])
+	must.Eq(t, "my-value", labels["my_key"])
+	must.MapNotContainsKey(t, labels, "some_dynamic_value")
+}
+
 // TestTaskRunner_IdentityHook_Enabled asserts that the identity hook exposes a
 // workload identity to a task.
 func TestTaskRunner_IdentityHook_Enabled(t *testing.T) {

client/config/config.go

@@ -203,6 +203,14 @@ type Config struct {
 	// allocation metrics to remote Telemetry sinks
 	PublishAllocationMetrics bool
 
+	// IncludeAllocMetadataInMetrics determines whether nomad should include the
+	// allocation metadata as labels in the metrics to remote Telemetry sinks
+	IncludeAllocMetadataInMetrics bool
+
+	// AllowedMetadataKeysInMetrics when provided nomad will only include the
+	// configured metadata keys as part of the metrics to remote Telemetry sinks
+	AllowedMetadataKeysInMetrics []string
+
 	// TLSConfig holds various TLS related configurations
 	TLSConfig *structsc.TLSConfig
 

command/agent/agent.go

@@ -861,6 +861,8 @@ func convertClientConfig(agentConfig *Config) (*clientconfig.Config, error) {
 	conf.StatsCollectionInterval = agentConfig.Telemetry.collectionInterval
 	conf.PublishNodeMetrics = agentConfig.Telemetry.PublishNodeMetrics
 	conf.PublishAllocationMetrics = agentConfig.Telemetry.PublishAllocationMetrics
+	conf.IncludeAllocMetadataInMetrics = agentConfig.Telemetry.IncludeAllocMetadataInMetrics
+	conf.AllowedMetadataKeysInMetrics = agentConfig.Telemetry.AllowedMetadataKeysInMetrics
 
 	// Set the TLS related configs
 	conf.TLSConfig = agentConfig.TLSConfig

command/agent/config.go

@@ -969,17 +969,19 @@ type Telemetry struct {
 	InMemoryRetentionPeriod string        `hcl:"in_memory_retention_period"`
 	inMemoryRetentionPeriod time.Duration `hcl:"-"`
 
-	StatsiteAddr             string        `hcl:"statsite_address"`
+	StatsiteAddr                  string        `hcl:"statsite_address"`
-	StatsdAddr               string        `hcl:"statsd_address"`
+	StatsdAddr                    string        `hcl:"statsd_address"`
-	DataDogAddr              string        `hcl:"datadog_address"`
+	DataDogAddr                   string        `hcl:"datadog_address"`
-	DataDogTags              []string      `hcl:"datadog_tags"`
+	DataDogTags                   []string      `hcl:"datadog_tags"`
-	PrometheusMetrics        bool          `hcl:"prometheus_metrics"`
+	PrometheusMetrics             bool          `hcl:"prometheus_metrics"`
-	DisableHostname          bool          `hcl:"disable_hostname"`
+	DisableHostname               bool          `hcl:"disable_hostname"`
-	UseNodeName              bool          `hcl:"use_node_name"`
+	UseNodeName                   bool          `hcl:"use_node_name"`
-	CollectionInterval       string        `hcl:"collection_interval"`
+	CollectionInterval            string        `hcl:"collection_interval"`
-	collectionInterval       time.Duration `hcl:"-"`
+	collectionInterval            time.Duration `hcl:"-"`
-	PublishAllocationMetrics bool          `hcl:"publish_allocation_metrics"`
+	PublishAllocationMetrics      bool          `hcl:"publish_allocation_metrics"`
-	PublishNodeMetrics       bool          `hcl:"publish_node_metrics"`
+	PublishNodeMetrics            bool          `hcl:"publish_node_metrics"`
+	IncludeAllocMetadataInMetrics bool          `hcl:"include_alloc_metadata_in_metrics"`
+	AllowedMetadataKeysInMetrics  []string      `hcl:"allowed_metadata_keys_in_metrics"`
 
 	// PrefixFilter allows for filtering out metrics from being collected
 	PrefixFilter []string `hcl:"prefix_filter"`
@@ -1343,6 +1345,8 @@ func DevConfig(mode *devModeConfig) *Config {
 	conf.Telemetry.PrometheusMetrics = true
 	conf.Telemetry.PublishAllocationMetrics = true
 	conf.Telemetry.PublishNodeMetrics = true
+	conf.Telemetry.IncludeAllocMetadataInMetrics = true
+	conf.Telemetry.AllowedMetadataKeysInMetrics = []string{}
 
 	if mode.consulMode {
 		conf.Consuls[0].ServiceIdentity = &config.WorkloadIdentityConfig{
@@ -2524,6 +2528,10 @@ func (t *Telemetry) Merge(b *Telemetry) *Telemetry {
 	if b.PublishAllocationMetrics {
 		result.PublishAllocationMetrics = true
 	}
+	if b.IncludeAllocMetadataInMetrics {
+		result.IncludeAllocMetadataInMetrics = true
+	}
+	result.AllowedMetadataKeysInMetrics = append(result.AllowedMetadataKeysInMetrics, b.AllowedMetadataKeysInMetrics...)
 	if b.CirconusAPIToken != "" {
 		result.CirconusAPIToken = b.CirconusAPIToken
 	}

website/content/docs/configuration/telemetry.mdx

@@ -58,6 +58,14 @@ The following options are available on all telemetry configurations.
 - `publish_allocation_metrics` `(bool: false)` - Specifies if Nomad should
   publish runtime metrics of allocations.
 
+- `include_alloc_metadata_in_metrics` `(bool: false)` - This controls whether
+  allocation metadata is included in metric labels. Enabling this option may result in
+  high cardinality labels. You should also configure [allowed_metadata_keys_in_metrics](#allowed_metadata_keys_in_metrics).
+
+- `allowed_metadata_keys_in_metrics` `(list: [])` - This filters the metadata
+  keys to be included in the metric publishing. By default it does not filter
+  out any keys and thus include all metadata.
+
 - `publish_node_metrics` `(bool: false)` - Specifies if Nomad should publish
   runtime metrics of nodes.