add collectorTargetReloadInterval (apply review)

bohub12 · bohub12 · commit ab9988c6fc7e · 2025-08-02T02:14:13.000+09:00
diff --git a/.chloggen/feature_add-collector-target-reload-interval.yaml b/.chloggen/feature_add-collector-target-reload-interval.yaml
@@ -0,0 +1,16 @@
+# One of 'breaking', 'deprecation', 'new_component', 'enhancement', 'bug_fix'
+change_type: enhancement
+
+# The name of the component, or a single word describing the area of concern, (e.g. collector, target allocator, auto-instrumentation, opamp, github action)
+component: target allocator
+
+# A brief description of the change. Surround your text with quotes ("") if it needs to start with a backtick (`).
+note: Add CollectorTargetReloadInterval option to configure the interval at which the Prometheus receiver reloads targets from the target allocator
+
+# One or more tracking issues related to the change
+issues: [4095]
+
+# (Optional) One or more lines of additional information to render under the primary note.
+# These lines will be padded with 2 spaces and then inserted directly into the document.
+# Use pipe (|) for multiline entries.
+subtext:
diff --git a/apis/v1beta1/opentelemetrycollector_types.go b/apis/v1beta1/opentelemetrycollector_types.go
@@ -227,6 +227,13 @@ type TargetAllocatorEmbedded struct {
 	// +kubebuilder:default:="30s"
 	// +kubebuilder:validation:Format:=duration
 	CollectorNotReadyGracePeriod *metav1.Duration `json:"collectorNotReadyGracePeriod,omitempty"`
+	// CollectorTargetReloadInterval defines the interval at which the Prometheus receiver will reload targets from the target allocator.
+	// The default is 30s.
+	//
+	// +optional
+	// +kubebuilder:default:="30s"
+	// +kubebuilder:validation:Format:=duration
+	CollectorTargetReloadInterval *metav1.Duration `json:"collectorTargetReloadInterval,omitempty"`
 }
 
 // Probe defines the OpenTelemetry's pod probe config.
diff --git a/apis/v1beta1/zz_generated.deepcopy.go b/apis/v1beta1/zz_generated.deepcopy.go
diff --git a/bundle/community/manifests/opentelemetry-operator.clusterserviceversion.yaml b/bundle/community/manifests/opentelemetry-operator.clusterserviceversion.yaml
@@ -99,7 +99,7 @@ metadata:
     categories: Logging & Tracing,Monitoring
     certified: "false"
     containerImage: ghcr.io/open-telemetry/opentelemetry-operator/opentelemetry-operator
-    createdAt: "2025-07-15T11:00:03Z"
+    createdAt: "2025-08-01T17:09:57Z"
     description: Provides the OpenTelemetry components, including the Collector
     operators.operatorframework.io/builder: operator-sdk-v1.29.0
     operators.operatorframework.io/project_layout: go.kubebuilder.io/v3
diff --git a/bundle/community/manifests/opentelemetry.io_opentelemetrycollectors.yaml b/bundle/community/manifests/opentelemetry.io_opentelemetrycollectors.yaml
@@ -7719,6 +7719,10 @@ spec:
                     default: 30s
                     format: duration
                     type: string
+                  collectorTargetReloadInterval:
+                    default: 30s
+                    format: duration
+                    type: string
                   enabled:
                     type: boolean
                   env:
diff --git a/bundle/openshift/manifests/opentelemetry-operator.clusterserviceversion.yaml b/bundle/openshift/manifests/opentelemetry-operator.clusterserviceversion.yaml
@@ -99,7 +99,7 @@ metadata:
     categories: Logging & Tracing,Monitoring
     certified: "false"
     containerImage: ghcr.io/open-telemetry/opentelemetry-operator/opentelemetry-operator
-    createdAt: "2025-07-15T11:00:04Z"
+    createdAt: "2025-08-01T17:09:57Z"
     description: Provides the OpenTelemetry components, including the Collector
     operators.operatorframework.io/builder: operator-sdk-v1.29.0
     operators.operatorframework.io/project_layout: go.kubebuilder.io/v3
diff --git a/bundle/openshift/manifests/opentelemetry.io_opentelemetrycollectors.yaml b/bundle/openshift/manifests/opentelemetry.io_opentelemetrycollectors.yaml
@@ -7719,6 +7719,10 @@ spec:
                     default: 30s
                     format: duration
                     type: string
+                  collectorTargetReloadInterval:
+                    default: 30s
+                    format: duration
+                    type: string
                   enabled:
                     type: boolean
                   env:
diff --git a/config/crd/bases/opentelemetry.io_opentelemetrycollectors.yaml b/config/crd/bases/opentelemetry.io_opentelemetrycollectors.yaml
@@ -7705,6 +7705,10 @@ spec:
                     default: 30s
                     format: duration
                     type: string
+                  collectorTargetReloadInterval:
+                    default: 30s
+                    format: duration
+                    type: string
                   enabled:
                     type: boolean
                   env:
diff --git a/docs/api/opentelemetrycollectors.md b/docs/api/opentelemetrycollectors.md
@@ -30747,6 +30747,17 @@ The default is 30s, which means that if a collector becomes not Ready, the targe
             <i>Default</i>: 30s<br/>
         </td>
         <td>false</td>
+      </tr><tr>
+        <td><b>collectorTargetReloadInterval</b></td>
+        <td>string</td>
+        <td>
+          CollectorTargetReloadInterval defines the interval at which the Prometheus receiver will reload targets from the target allocator.
+The default is 30s.<br/>
+          <br/>
+            <i>Format</i>: duration<br/>
+            <i>Default</i>: 30s<br/>
+        </td>
+        <td>false</td>
       </tr><tr>
         <td><b>enabled</b></td>
         <td>boolean</td>
diff --git a/internal/manifests/collector/config_replace.go b/internal/manifests/collector/config_replace.go
@@ -58,6 +58,12 @@ func ReplaceConfig(otelcol v1beta1.OpenTelemetryCollector, targetAllocator *v1al
 		return "", validateCfgPromErr
 	}
 
+	// Use the interval from CRD (which has a default of 30s if not specified)
+	if otelcol.Spec.TargetAllocator.CollectorTargetReloadInterval != nil {
+		interval := otelcol.Spec.TargetAllocator.CollectorTargetReloadInterval.Duration.String()
+		options = append(options, ta.WithCollectorTargetReloadInterval(interval))
+	}
+
 	// To avoid issues caused by Prometheus validation logic, which fails regex validation when it encounters
 	// $$ in the prom config, we update the YAML file directly without marshaling and unmarshalling.
 	updPromCfgMap, getCfgPromErr := ta.AddTAConfigToPromConfig(promCfgMap, naming.TAService(targetAllocator.Name), options...)
diff --git a/internal/manifests/collector/config_replace_test.go b/internal/manifests/collector/config_replace_test.go
@@ -6,10 +6,12 @@ package collector
 import (
 	"os"
 	"testing"
+	"time"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 	"gopkg.in/yaml.v2"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 
 	ta "github.com/open-telemetry/opentelemetry-operator/internal/manifests/targetallocator/adapters"
 )
@@ -125,4 +127,18 @@ func TestReplaceConfig(t *testing.T) {
 
 		assert.YAMLEq(t, expectedConfig, actualConfig)
 	})
+
+	t.Run("should update collectorTargetReloadInterval if specified", func(t *testing.T) {
+		customInterval := &metav1.Duration{Duration: 10 * time.Second}
+
+		param.OtelCol.Spec.TargetAllocator.CollectorTargetReloadInterval = customInterval
+
+		actualConfig, err := ReplaceConfig(param.OtelCol, param.TargetAllocator, ta.WithCollectorTargetReloadInterval(customInterval.Duration.String()))
+		assert.NoError(t, err)
+
+		promCfgMap, err := ta.ConfigToPromConfig(actualConfig)
+		assert.NoError(t, err)
+
+		assert.Equal(t, customInterval.Duration.String(), promCfgMap["target_allocator"].(map[interface{}]interface{})["interval"])
+	})
 }
diff --git a/internal/manifests/targetallocator/adapters/config_to_prom_config.go b/internal/manifests/targetallocator/adapters/config_to_prom_config.go
@@ -269,6 +269,14 @@ func WithTLSConfig(caFile, certFile, keyFile, taServiceName string) TAOption {
 	}
 }
 
+// WithCollectorTargetReloadInterval adds the collector target reload interval to the target allocator configuration.
+func WithCollectorTargetReloadInterval(interval string) TAOption {
+	return func(targetAllocatorCfg map[interface{}]interface{}) error {
+		targetAllocatorCfg["interval"] = interval
+		return nil
+	}
+}
+
 // AddTAConfigToPromConfig adds or updates the target_allocator configuration in the Prometheus configuration.
 // If the `EnableTargetAllocatorRewrite` feature flag for the target allocator is enabled, this function
 // removes the existing scrape_configs from the collector's Prometheus configuration as it's not required.
diff --git a/internal/manifests/targetallocator/adapters/config_to_prom_config_test.go b/internal/manifests/targetallocator/adapters/config_to_prom_config_test.go
@@ -549,3 +549,40 @@ func TestAddTAConfigToPromConfigWithTLSConfig(t *testing.T) {
 		assert.Equal(t, expectedResult, result)
 	})
 }
+
+func TestAddTAConfigToPromConfigWithCollectorTargetReloadInterval(t *testing.T) {
+	t.Run("should return expected prom config map with TA config and collector target reload interval", func(t *testing.T) {
+		cfg := map[interface{}]interface{}{
+			"config": map[interface{}]interface{}{
+				"scrape_configs": []interface{}{
+					map[interface{}]interface{}{
+						"job_name": "test_job",
+						"static_configs": []interface{}{
+							map[interface{}]interface{}{
+								"targets": []interface{}{
+									"localhost:9090",
+								},
+							},
+						},
+					},
+				},
+			},
+		}
+
+		taServiceName := "test-targetallocator"
+
+		expectedResult := map[interface{}]interface{}{
+			"config": map[interface{}]interface{}{},
+			"target_allocator": map[interface{}]interface{}{
+				"endpoint":     "http://test-targetallocator:80",
+				"interval":     "10s",
+				"collector_id": "${POD_NAME}",
+			},
+		}
+
+		result, err := ta.AddTAConfigToPromConfig(cfg, taServiceName, ta.WithCollectorTargetReloadInterval("10s"))
+
+		assert.NoError(t, err)
+		assert.Equal(t, expectedResult, result)
+	})
+}
