x-pack/metricbeat/module/gcp: Add omitted labels (elastic#40062)

* gcp ommitted label fix

* added comments

* resolved comments

* CHANGELOG.next.asciidoc

CHANGELOG.next.asciidoc

@@ -179,6 +179,7 @@ https://github.com/elastic/beats/compare/v8.8.1\...main[Check the HEAD diff]
 - Fix behavior of pagetypeinfo metrics {pull}39985[39985]
 - Fix query logic for temp and non-temp tablespaces in Oracle module. {issue}38051[38051] {pull}39787[39787]
 - Set GCP metrics config period to the default (60s) when the value is below the minimum allowed period. {issue}30434[30434] {pull}40020[40020]
+- Add GCP 'instance_id' resource label in ECS cloud fields. {issue}40033[40033] {pull}40062[40062]
 - Fix missing metrics from CloudWatch when include_linked_accounts set to false. {issue}40071[40071] {pull}40135[40135]
 - Update beat module with apm-server monitoring metrics fields {pull}40127[40127]
 

x-pack/metricbeat/module/gcp/timeseries_metadata_collector.go

@@ -100,7 +100,15 @@ func (s *StackdriverTimeSeriesMetadataCollector) Metadata(ctx context.Context, i
 		// common.Mapstr seems to not work as expected when deleting keys so I have to iterate over all results to add
 		// the ones I want
 		for k, v := range s.timeSeries.Resource.Labels {
-			if k == TimeSeriesResponsePathForECSAvailabilityZone || k == TimeSeriesResponsePathForECSInstanceID || k == TimeSeriesResponsePathForECSAccountID {
+
+			// We are omitting some labels here because they are added separately for services with additional metadata logic.
+			// However, we explicitly include the instance_id label to ensure it is not missed for services without additional metadata logic.
+			if k == TimeSeriesResponsePathForECSInstanceID {
+				_, _ = ecs.Put(ECSCloudInstanceIDKey, v)
+				continue
+			}
+
+			if k == TimeSeriesResponsePathForECSAvailabilityZone || k == TimeSeriesResponsePathForECSAccountID {
 				continue
 			}