[receiver/vcenter] Adds vSAN metrics for Virtual Machines

.chloggen/vcenterreceiver-vm-vsan.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: 'enhancement'
+
+# The name of the component, or a single word describing the area of concern, (e.g. filelogreceiver)
+component: 'vcenterreceiver'
+
+# A brief description of the change.  Surround your text with quotes ("") if it needs to start with a backtick (`).
+note: Adds a number of default disabled vSAN metrics for Virtual Machines.
+
+# Mandatory: One or more tracking issues related to the change. You can use the PR number here if no issue exists.
+issues: [33556]
+
+# (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: []

.gitignore

@@ -26,6 +26,7 @@ dist/
 # Miscellaneous files
 *.sw[op]
 *.DS_Store
+__debug_bin*
 
 # Coverage
 coverage/*

receiver/vcenterreceiver/client.go

@@ -7,35 +7,45 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"maps"
 	"net/url"
+	"reflect"
+	"strconv"
+	"strings"
+	"time"
 
 	"github.com/vmware/govmomi"
 	"github.com/vmware/govmomi/find"
 	"github.com/vmware/govmomi/object"
 	"github.com/vmware/govmomi/performance"
-	"github.com/vmware/govmomi/property"
 	"github.com/vmware/govmomi/view"
 	"github.com/vmware/govmomi/vim25"
 	"github.com/vmware/govmomi/vim25/mo"
+	"github.com/vmware/govmomi/vim25/soap"
 	vt "github.com/vmware/govmomi/vim25/types"
+	"github.com/vmware/govmomi/vsan"
+	"github.com/vmware/govmomi/vsan/types"
+	"go.uber.org/zap"
 )
 
 // vcenterClient is a client that collects data from a vCenter endpoint.
 type vcenterClient struct {
-	moClient  *govmomi.Client
-	vimDriver *vim25.Client
-	finder    *find.Finder
-	pc        *property.Collector
-	pm        *performance.Manager
-	vm        *view.Manager
-	cfg       *Config
+	logger     *zap.Logger
+	moClient   *govmomi.Client
+	vimDriver  *vim25.Client
+	vsanDriver *vsan.Client
+	finder     *find.Finder
+	pm         *performance.Manager
+	vm         *view.Manager
+	cfg        *Config
 }
 
 var newVcenterClient = defaultNewVcenterClient
 
-func defaultNewVcenterClient(c *Config) *vcenterClient {
+func defaultNewVcenterClient(l *zap.Logger, c *Config) *vcenterClient {
 	return &vcenterClient{
-		cfg: c,
+		logger: l,
+		cfg:    c,
 	}
 }
 
@@ -70,10 +80,15 @@ func (vc *vcenterClient) EnsureConnection(ctx context.Context) error {
 	}
 	vc.moClient = client
 	vc.vimDriver = client.Client
-	vc.pc = property.DefaultCollector(vc.vimDriver)
 	vc.finder = find.NewFinder(vc.vimDriver)
 	vc.pm = performance.NewManager(vc.vimDriver)
 	vc.vm = view.NewManager(vc.vimDriver)
+	vsanDriver, err := vsan.NewClient(ctx, vc.vimDriver)
+	if err != nil {
+		vc.logger.Info(fmt.Errorf("could not create VSAN client: %w", err).Error())
+	} else {
+		vc.vsanDriver = vsanDriver
+	}
 	return nil
 }
 
@@ -320,3 +335,240 @@ func (vc *vcenterClient) PerfMetricsQuery(
 		resultsByRef: resultsByRef,
 	}, nil
 }
+
+// VSANQueryResults contains all returned vSAN metric related data
+type VSANQueryResults struct {
+	// Contains vSAN metric data keyed by UUID string
+	MetricResultsByUUID map[string]*VSANMetricResults
+}
+
+// VSANMetricResults contains vSAN metric related data for a single resource
+type VSANMetricResults struct {
+	// Contains UUID info for related resource
+	UUID string
+	// Contains returned metric value info for all metrics
+	MetricDetails []*VSANMetricDetails
+}
+
+// VSANMetricDetails contains vSAN metric data for a single metric
+type VSANMetricDetails struct {
+	// Contains the metric label
+	MetricLabel string
+	// Contains the metric interval in seconds
+	Interval int32
+	// Contains timestamps for all metric values
+	Timestamps []*time.Time
+	// Contains all values for vSAN metric label
+	Values []int64
+}
+
+// vSANQueryType represents the type of VSAN query
+type vSANQueryType string
+
+const (
+	VSANQueryTypeVirtualMachines vSANQueryType = "virtual-machine:*"
+)
+
+// getLabelsForQueryType returns the appropriate labels for each query type
+func (vc *vcenterClient) getLabelsForQueryType(queryType vSANQueryType) []string {
+	switch queryType {
+	case VSANQueryTypeVirtualMachines:
+		return []string{
+			"iopsRead", "iopsWrite", "throughputRead", "throughputWrite",
+			"latencyRead", "latencyWrite",
+		}
+	default:
+		return []string{}
+	}
+}
+
+// VSANVirtualMachines returns back virtual machine vSAN performance metrics
+func (vc *vcenterClient) VSANVirtualMachines(
+	ctx context.Context,
+	clusterRefs []*vt.ManagedObjectReference,
+) (*VSANQueryResults, error) {
+	results, err := vc.vSANQuery(ctx, VSANQueryTypeVirtualMachines, clusterRefs)
+	err = vc.handleVSANError(err, VSANQueryTypeVirtualMachines)
+	return results, err
+}
+
+// vSANQuery performs a vSAN query for the specified type across all clusters
+func (vc *vcenterClient) vSANQuery(
+	ctx context.Context,
+	queryType vSANQueryType,
+	clusterRefs []*vt.ManagedObjectReference,
+) (*VSANQueryResults, error) {
+	allResults := VSANQueryResults{
+		MetricResultsByUUID: map[string]*VSANMetricResults{},
+	}
+
+	for _, clusterRef := range clusterRefs {
+		results, err := vc.vSANQueryByCluster(ctx, queryType, clusterRef)
+		if err != nil {
+			return &allResults, err
+		}
+
+		maps.Copy(allResults.MetricResultsByUUID, results.MetricResultsByUUID)
+	}
+
+	return &allResults, nil
+}
+
+// vSANQueryByCluster performs a vSAN query for the specified type for one cluster
+func (vc *vcenterClient) vSANQueryByCluster(
+	ctx context.Context,
+	queryType vSANQueryType,
+	clusterRef *vt.ManagedObjectReference,
+) (*VSANQueryResults, error) {
+	queryResults := VSANQueryResults{
+		MetricResultsByUUID: map[string]*VSANMetricResults{},
+	}
+	// Not all vCenters support vSAN so just return an empty result
+	if vc.vsanDriver == nil {
+		return &queryResults, nil
+	}
+
+	now := time.Now()
+	querySpec := []types.VsanPerfQuerySpec{
+		{
+			EntityRefId: string(queryType),
+			StartTime:   &now,
+			EndTime:     &now,
+			Labels:      vc.getLabelsForQueryType(queryType),
+		},
+	}
+	rawResults, err := vc.vsanDriver.VsanPerfQueryPerf(ctx, clusterRef, querySpec)
+	if err != nil {
+		return nil, fmt.Errorf("problem retrieving %s vSAN metrics for cluster %s: %w", queryType, clusterRef.Value, err)
+	}
+
+	queryResults.MetricResultsByUUID = map[string]*VSANMetricResults{}
+	for _, rawResult := range rawResults {
+		metricResults, err := vc.convertVSANResultToMetricResults(rawResult)
+		if err != nil && metricResults != nil {
+			return &queryResults, fmt.Errorf("problem processing %s [%s] vSAN metrics for cluster %s: %w", queryType, metricResults.UUID, clusterRef.Value, err)
+		}
+		if err != nil {
+			return &queryResults, fmt.Errorf("problem processing %s vSAN metrics for cluster %s: %w", queryType, clusterRef.Value, err)
+		}
+
+		queryResults.MetricResultsByUUID[metricResults.UUID] = metricResults
+	}
+	return &queryResults, nil
+}
+
+func (vc *vcenterClient) handleVSANError(
+	err error,
+	queryType vSANQueryType,
+) error {
+	faultErr := errors.Unwrap(err)
+	if faultErr == nil {
+		return err
+	}
+	if !soap.IsSoapFault(faultErr) {
+		return err
+	}
+
+	fault := soap.ToSoapFault(faultErr)
+	msg := fault.String
+
+	if fault.Detail.Fault != nil {
+		msg = reflect.TypeOf(fault.Detail.Fault).Name()
+	}
+	switch msg {
+	case "NotSupported":
+		vc.logger.Debug(fmt.Sprintf("%s vSAN metrics not supported: %s", queryType, err.Error()))
+		return nil
+	case "NotFound":
+		vc.logger.Debug(fmt.Sprintf("no %s vSAN metrics found: %s", queryType, err.Error()))
+		return nil
+	default:
+		return err
+	}
+}
+
+func (vc *vcenterClient) convertVSANResultToMetricResults(vSANResult types.VsanPerfEntityMetricCSV) (*VSANMetricResults, error) {
+	uuid, err := vc.uuidFromEntityRefID(vSANResult.EntityRefId)
+	if err != nil {
+		return nil, err
+	}
+
+	metricResults := VSANMetricResults{
+		UUID:          uuid,
+		MetricDetails: []*VSANMetricDetails{},
+	}
+
+	// Parse all timestamps
+	localZone, _ := time.Now().Local().Zone()
+	timeStrings := strings.Split(vSANResult.SampleInfo, ",")
+	timestamps := []time.Time{}
+	for _, timeString := range timeStrings {
+		// Assuming the collector is making the request in the same time zone as the localized response
+		// from the vSAN API. Not a great assumption, but otherwise it will almost definitely be wrong
+		// if we assume that it is UTC. There is precedent for this method at least.
+		timestamp, err := time.Parse("2006-01-02 15:04:05 MST", fmt.Sprintf("%s %s", timeString, localZone))
+		if err != nil {
+			return &metricResults, fmt.Errorf("problem parsing timestamp from %s: %w", timeString, err)
+		}
+
+		timestamps = append(timestamps, timestamp)
+	}
+
+	// Parse all metrics
+	for _, vSANValue := range vSANResult.Value {
+		metricDetails, err := vc.convertVSANValueToMetricDetails(vSANValue, timestamps)
+		if err != nil {
+			return &metricResults, err
+		}
+
+		metricResults.MetricDetails = append(metricResults.MetricDetails, metricDetails)
+	}
+	return &metricResults, nil
+}
+
+func (vc *vcenterClient) convertVSANValueToMetricDetails(
+	vSANValue types.VsanPerfMetricSeriesCSV,
+	timestamps []time.Time,
+) (*VSANMetricDetails, error) {
+	metricLabel := vSANValue.MetricId.Label
+	metricInterval := vSANValue.MetricId.MetricsCollectInterval
+	// If not found assume the interval is 5m
+	if metricInterval == 0 {
+		vc.logger.Warn(fmt.Sprintf("no interval found for vSAN metric [%s] so assuming 5m", metricLabel))
+		metricInterval = 300
+	}
+	metricDetails := VSANMetricDetails{
+		MetricLabel: metricLabel,
+		Interval:    metricInterval,
+		Timestamps:  []*time.Time{},
+		Values:      []int64{},
+	}
+	valueStrings := strings.Split(vSANValue.Values, ",")
+	if len(valueStrings) != len(timestamps) {
+		return nil, fmt.Errorf("number of timestamps [%d] doesn't match number of values [%d] for metric %s", len(timestamps), len(valueStrings), metricLabel)
+	}
+
+	// Match up timestamps with metric values
+	for i, valueString := range valueStrings {
+		value, err := strconv.ParseInt(valueString, 10, 64)
+		if err != nil {
+			return nil, fmt.Errorf("problem converting value [%s] for metric %s", valueString, metricLabel)
+		}
+
+		metricDetails.Timestamps = append(metricDetails.Timestamps, &timestamps[i])
+		metricDetails.Values = append(metricDetails.Values, value)
+	}
+
+	return &metricDetails, nil
+}
+
+// uuidFromEntityRefID returns the UUID portion of the EntityRefId
+func (vc *vcenterClient) uuidFromEntityRefID(id string) (string, error) {
+	colonIndex := strings.Index(id, ":")
+	if colonIndex != -1 {
+		uuid := id[colonIndex+1:]
+		return uuid, nil
+	}
+
+	return "", fmt.Errorf("no ':' found in EntityRefId [%s] to parse UUID", id)
+}