Fix: Skip vSAN collection and logging when all vSAN metrics are disabled (open-telemetry#43704)

#### Description
This PR adds a new method in the `vcenter` receiver component called
`hasEnabledVSANMetrics()`. Whether or not vSAN metrics are enabled is
calculated by checking if all metrics are disabled, if any are enabled
then the method will return `true`, else it will return `false`.

The method is then used to guard vSAN-related initialization and logging
logic - when all vSAN metrics are disabled, the receiver will now skip
vSAN API setup and suppress related warning or error logs.

#### Link to tracking issue

open-telemetry#38489

#### Testing
I've added `metadata_test.go`, which uses the `metadata` file itself to
detect which vSAN metrics are available, and will fail the test if any
are added or removed. The test will pass if what is returned from
`getCheckedVSANMetrics()` matches what is in `hasEnabledVSANMetrics()`.
This does add an extra weird step for developers editing the metadata
yaml, since the test basically instructs them to add or remove a metric
from two places: the test and the source file. It's a bit redundant, but
feels clear enough for now

Maybe there's a better way to achieve this? I also played around with a
reflection based approach in `hasEnabledVSANMetrics` but this felt a bit
icky. Open to feedback on that one :)

#### Documentation

---------

Co-authored-by: Keith Schmitt <[email protected]>
Co-authored-by: Paulo Janotti <[email protected]>

Author: 3 people

.chloggen/blewis_only-collect-vsan-metrics-if-enabled.yaml

@@ -0,0 +1,27 @@
+# Use this changelog template to create an entry for release notes.
+
+# One of 'breaking', 'deprecation', 'new_component', 'enhancement', 'bug_fix'
+change_type: 'bug_fix'
+
+# The name of the component, or a single word describing the area of concern, (e.g. receiver/filelog)
+component: 'receiver/vcenter'
+
+# A brief description of the change.  Surround your text with quotes ("") if it needs to start with a backtick (`).
+note: 'Skip vSAN collection and logging when all vSAN metrics are disabled'
+
+# Mandatory: One or more tracking issues related to the change. You can use the PR number here if no issue exists.
+issues: [38489]
+
+# (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:
+
+# If your change doesn't affect end users or the exported elements of any package,
+# you should instead start your pull request title with [chore] or use the "Skip Changelog" label.
+# Optional: The change log or logs in which this entry should be included.
+# e.g. '[user]' or '[user, api]'
+# Include 'user' if the change is relevant to end users.
+# Include 'api' if there is a change to a library API.
+# Default: '[user]'
+change_logs: [api]

receiver/vcenterreceiver/go.mod

@@ -27,6 +27,7 @@ require (
 	go.uber.org/goleak v1.3.0
 	go.uber.org/multierr v1.11.0
 	go.uber.org/zap v1.27.0
+	gopkg.in/yaml.v3 v3.0.1
 )
 
 require (
@@ -114,7 +115,6 @@ require (
 	google.golang.org/genproto/googleapis/rpc v0.0.0-20250804133106-a7a43d27e69b // indirect
 	google.golang.org/grpc v1.76.0 // indirect
 	google.golang.org/protobuf v1.36.10 // indirect
-	gopkg.in/yaml.v3 v3.0.1 // indirect
 )
 
 replace github.com/open-telemetry/opentelemetry-collector-contrib/pkg/pdatatest => ../../pkg/pdatatest

receiver/vcenterreceiver/metadata_test.go

@@ -0,0 +1,134 @@
+// Copyright The OpenTelemetry Authors
+// SPDX-License-Identifier: Apache-2.0
+
+package vcenterreceiver
+
+import (
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"gopkg.in/yaml.v3"
+)
+
+// TestVSANMetricsFromMetadata ensures hasEnabledVSANMetrics() covers all vSAN metrics
+// defined in metadata.yaml. This test will fail if a new vSAN metric is added to
+// metadata.yaml but not included in hasEnabledVSANMetrics().
+func TestVSANMetricsFromMetadata(t *testing.T) {
+	vSANMetricsFromYAML := getVSANMetricsFromMetadataYAML(t)
+	checkedMetrics := getCheckedVSANMetrics()
+
+	// Convert to sets for comparison
+	yamlSet := make(map[string]bool)
+	for _, metric := range vSANMetricsFromYAML {
+		yamlSet[metric] = true
+	}
+
+	checkedSet := make(map[string]bool)
+	for _, metric := range checkedMetrics {
+		checkedSet[metric] = true
+	}
+
+	var missingInFunction []string
+	for metric := range yamlSet {
+		if !checkedSet[metric] {
+			missingInFunction = append(missingInFunction, metric)
+		}
+	}
+
+	var extraInFunction []string
+	for metric := range checkedSet {
+		if !yamlSet[metric] {
+			extraInFunction = append(extraInFunction, metric)
+		}
+	}
+
+	if len(missingInFunction) > 0 {
+		t.Errorf("hasEnabledVSANMetrics() is missing these vSAN metrics from metadata.yaml:")
+		for _, metric := range missingInFunction {
+			t.Errorf("   - %s", metric)
+		}
+		t.Error("Please add the missing metrics to hasEnabledVSANMetrics() function")
+	}
+
+	if len(extraInFunction) > 0 {
+		t.Errorf("hasEnabledVSANMetrics() checks these metrics not found in metadata.yaml:")
+		for _, metric := range extraInFunction {
+			t.Errorf("   - %s", metric)
+		}
+		t.Error("Please remove non-vSAN metrics from hasEnabledVSANMetrics() function")
+	}
+
+	if len(missingInFunction) == 0 && len(extraInFunction) == 0 {
+		t.Logf("All %d vSAN metrics from metadata.yaml are properly checked", len(vSANMetricsFromYAML))
+	}
+}
+
+type metadataYAML struct {
+	Metrics map[string]any `yaml:"metrics"`
+}
+
+func getVSANMetricsFromMetadataYAML(t *testing.T) []string {
+	metadataPath := findMetadataYAML(t)
+
+	data, err := os.ReadFile(metadataPath)
+	if err != nil {
+		t.Fatalf("Failed to read metadata.yaml: %v", err)
+	}
+
+	var metadata metadataYAML
+	if err := yaml.Unmarshal(data, &metadata); err != nil {
+		t.Fatalf("Failed to parse metadata.yaml: %v", err)
+	}
+
+	// Extract vSAN metrics
+	var vSANMetrics []string
+	for metricName := range metadata.Metrics {
+		if strings.Contains(metricName, "vsan") {
+			vSANMetrics = append(vSANMetrics, metricName)
+		}
+	}
+
+	return vSANMetrics
+}
+
+func findMetadataYAML(t *testing.T) string {
+	dir, err := os.Getwd()
+	if err != nil {
+		t.Fatalf("Failed to get current directory: %v", err)
+	}
+
+	// Look for metadata.yaml in current directory
+	metadataPath := filepath.Join(dir, "metadata.yaml")
+	if _, err := os.Stat(metadataPath); err == nil {
+		return metadataPath
+	}
+
+	t.Fatalf("Could not find metadata.yaml")
+	return ""
+}
+
+// getCheckedVSANMetrics returns the metrics checked in hasEnabledVSANMetrics()
+func getCheckedVSANMetrics() []string {
+	// This should match exactly what's in hasEnabledVSANMetrics()
+	return []string{
+		// Cluster vSAN metrics
+		"vcenter.cluster.vsan.congestions",
+		"vcenter.cluster.vsan.latency.avg",
+		"vcenter.cluster.vsan.operations",
+		"vcenter.cluster.vsan.throughput",
+
+		// Host vSAN metrics
+		"vcenter.host.vsan.cache.hit_rate",
+		"vcenter.host.vsan.congestions",
+		"vcenter.host.vsan.latency.avg",
+		"vcenter.host.vsan.operations",
+		"vcenter.host.vsan.throughput",
+
+		// VM vSAN metrics
+		"vcenter.vm.vsan.latency.avg",
+		"vcenter.vm.vsan.operations",
+		"vcenter.vm.vsan.throughput",
+	}
+}

receiver/vcenterreceiver/processors.go

@@ -198,14 +198,16 @@ func (v *vcenterMetricScraper) buildHostMetrics(
 		v.recordHostPerformanceMetrics(hostPerfMetrics)
 	}
 
-	if hs.Config == nil || hs.Config.VsanHostConfig == nil || hs.Config.VsanHostConfig.ClusterInfo == nil {
-		v.logger.Info("couldn't determine UUID necessary for vSAN metrics for host " + hs.Name)
-		v.mb.EmitForResource(metadata.WithResource(rb.Emit()))
-		return vmRefToComputeRef, nil
-	}
-	vSANMetrics := v.scrapeData.hostVSANMetricsByUUID[hs.Config.VsanHostConfig.ClusterInfo.NodeUuid]
-	if vSANMetrics != nil {
-		v.recordHostVSANMetrics(vSANMetrics)
+	if v.hasEnabledVSANMetrics() {
+		if hs.Config == nil || hs.Config.VsanHostConfig == nil || hs.Config.VsanHostConfig.ClusterInfo == nil {
+			v.logger.Info("couldn't determine UUID necessary for vSAN metrics for host " + hs.Name)
+			v.mb.EmitForResource(metadata.WithResource(rb.Emit()))
+			return vmRefToComputeRef, nil
+		}
+		vSANMetrics := v.scrapeData.hostVSANMetricsByUUID[hs.Config.VsanHostConfig.ClusterInfo.NodeUuid]
+		if vSANMetrics != nil {
+			v.recordHostVSANMetrics(vSANMetrics)
+		}
 	}
 	v.mb.EmitForResource(metadata.WithResource(rb.Emit()))
 
@@ -332,9 +334,11 @@ func (v *vcenterMetricScraper) buildVMMetrics(
 		v.recordVMPerformanceMetrics(perfMetrics)
 	}
 
-	vSANMetrics := v.scrapeData.vmVSANMetricsByUUID[vm.Config.InstanceUuid]
-	if vSANMetrics != nil {
-		v.recordVMVSANMetrics(vSANMetrics)
+	if v.hasEnabledVSANMetrics() {
+		vSANMetrics := v.scrapeData.vmVSANMetricsByUUID[vm.Config.InstanceUuid]
+		if vSANMetrics != nil {
+			v.recordVMVSANMetrics(vSANMetrics)
+		}
 	}
 	v.mb.EmitForResource(metadata.WithResource(rb.Emit()))
 
@@ -380,16 +384,19 @@ func (v *vcenterMetricScraper) buildClusterMetrics(
 	}
 	// Record and emit Cluster metric data points
 	v.recordClusterStats(ts, cr, vmGroupInfo)
-	vSANConfig := cr.ConfigurationEx.(*types.ClusterConfigInfoEx).VsanConfigInfo
-	if vSANConfig == nil || vSANConfig.Enabled == nil || !*vSANConfig.Enabled || vSANConfig.DefaultConfig == nil {
-		v.logger.Info("couldn't determine UUID necessary for vSAN metrics for cluster " + cr.Name)
-		v.mb.EmitForResource(metadata.WithResource(rb.Emit()))
-		return err
-	}
 
-	vSANMetrics := v.scrapeData.clusterVSANMetricsByUUID[vSANConfig.DefaultConfig.Uuid]
-	if vSANMetrics != nil {
-		v.recordClusterVSANMetrics(vSANMetrics)
+	if v.hasEnabledVSANMetrics() {
+		vSANConfig := cr.ConfigurationEx.(*types.ClusterConfigInfoEx).VsanConfigInfo
+		if vSANConfig == nil || vSANConfig.Enabled == nil || !*vSANConfig.Enabled || vSANConfig.DefaultConfig == nil {
+			v.logger.Info("couldn't determine UUID necessary for vSAN metrics for cluster " + cr.Name)
+			v.mb.EmitForResource(metadata.WithResource(rb.Emit()))
+			return err
+		}
+
+		vSANMetrics := v.scrapeData.clusterVSANMetricsByUUID[vSANConfig.DefaultConfig.Uuid]
+		if vSANMetrics != nil {
+			v.recordClusterVSANMetrics(vSANMetrics)
+		}
 	}
 
 	v.mb.EmitForResource(metadata.WithResource(rb.Emit()))

receiver/vcenterreceiver/scraper.go

@@ -88,6 +88,36 @@ func newVcenterScrapeData() *vcenterScrapeData {
 	}
 }
 
+func (v *vcenterMetricScraper) hasEnabledVSANMetrics() bool {
+	metrics := v.config.Metrics
+
+	// Check cluster vSAN metrics
+	if metrics.VcenterClusterVsanCongestions.Enabled ||
+		metrics.VcenterClusterVsanLatencyAvg.Enabled ||
+		metrics.VcenterClusterVsanOperations.Enabled ||
+		metrics.VcenterClusterVsanThroughput.Enabled {
+		return true
+	}
+
+	// Check host vSAN metrics
+	if metrics.VcenterHostVsanCacheHitRate.Enabled ||
+		metrics.VcenterHostVsanCongestions.Enabled ||
+		metrics.VcenterHostVsanLatencyAvg.Enabled ||
+		metrics.VcenterHostVsanOperations.Enabled ||
+		metrics.VcenterHostVsanThroughput.Enabled {
+		return true
+	}
+
+	// Check VM vSAN metrics
+	if metrics.VcenterVMVsanLatencyAvg.Enabled ||
+		metrics.VcenterVMVsanOperations.Enabled ||
+		metrics.VcenterVMVsanThroughput.Enabled {
+		return true
+	}
+
+	return false
+}
+
 func (v *vcenterMetricScraper) Start(ctx context.Context, _ component.Host) error {
 	connectErr := v.client.EnsureConnection(ctx)
 	// don't fail to start if we cannot establish connection, just log an error
@@ -246,13 +276,15 @@ func (v *vcenterMetricScraper) scrapeComputes(ctx context.Context, dc *mo.Datace
 		}
 	}
 
-	// Get all Cluster vSAN metrics and store for later retrieval
-	vSANMetrics, err := v.client.VSANClusters(ctx, v.scrapeData.clusterRefs)
-	if err != nil {
-		errs.AddPartial(1, fmt.Errorf("failed to retrieve vSAN metrics for Clusters: %w", err))
-		return
+	if v.hasEnabledVSANMetrics() {
+		// Get all Cluster vSAN metrics and store for later retrieval (only if vSAN metrics are enabled)
+		vSANMetrics, err := v.client.VSANClusters(ctx, v.scrapeData.clusterRefs)
+		if err != nil {
+			errs.AddPartial(1, fmt.Errorf("failed to retrieve vSAN metrics for Clusters: %w", err))
+			return
+		}
+		v.scrapeData.clusterVSANMetricsByUUID = vSANMetrics.MetricResultsByUUID
 	}
-	v.scrapeData.clusterVSANMetricsByUUID = vSANMetrics.MetricResultsByUUID
 }
 
 // scrapeHosts scrapes and stores all relevant metric/property data for a Datacenter's HostSystems
@@ -287,12 +319,15 @@ func (v *vcenterMetricScraper) scrapeHosts(ctx context.Context, dc *mo.Datacente
 		v.scrapeData.hostPerfMetricsByRef = results.resultsByRef
 	}
 
-	vSANMetrics, err := v.client.VSANHosts(ctx, v.scrapeData.clusterRefs)
-	if err != nil {
-		errs.AddPartial(1, fmt.Errorf("failed to retrieve vSAN metrics for Hosts: %w", err))
-		return
+	if v.hasEnabledVSANMetrics() {
+		// Get all Host vSAN metrics and store for later retrieval
+		vSANMetrics, err := v.client.VSANHosts(ctx, v.scrapeData.clusterRefs)
+		if err != nil {
+			errs.AddPartial(1, fmt.Errorf("failed to retrieve vSAN metrics for Hosts: %w", err))
+			return
+		}
+		v.scrapeData.hostVSANMetricsByUUID = vSANMetrics.MetricResultsByUUID
 	}
-	v.scrapeData.hostVSANMetricsByUUID = vSANMetrics.MetricResultsByUUID
 }
 
 // scrapeResourcePools scrapes and stores all relevant property data for a Datacenter's ResourcePools/vApps
@@ -343,12 +378,13 @@ func (v *vcenterMetricScraper) scrapeVirtualMachines(ctx context.Context, dc *mo
 		v.scrapeData.vmPerfMetricsByRef = results.resultsByRef
 	}
 
-	// Get all VirtualMachine vSAN metrics and store for later retrieval
-	vSANMetrics, err := v.client.VSANVirtualMachines(ctx, v.scrapeData.clusterRefs)
-	if err != nil {
-		errs.AddPartial(1, fmt.Errorf("failed to retrieve vSAN metrics for VirtualMachines: %w", err))
-		return
+	if v.hasEnabledVSANMetrics() {
+		// Get all VirtualMachine vSAN metrics and store for later retrieval
+		vSANMetrics, err := v.client.VSANVirtualMachines(ctx, v.scrapeData.clusterRefs)
+		if err != nil {
+			errs.AddPartial(1, fmt.Errorf("failed to retrieve vSAN metrics for VirtualMachines: %w", err))
+			return
+		}
+		v.scrapeData.vmVSANMetricsByUUID = vSANMetrics.MetricResultsByUUID
 	}
-
-	v.scrapeData.vmVSANMetricsByUUID = vSANMetrics.MetricResultsByUUID
 }