system: refactor collector (#1730)

Signed-off-by: Jan-Otto Kröpke <[email protected]>

exporter.go

@@ -179,6 +179,12 @@ func run() int {
 
 	logger.Debug("Logging has Started")
 
+	if v, ok := os.LookupEnv("WINDOWS_EXPORTER_PERF_COUNTERS_ENGINE"); ok && v == "pdh" || *togglePDH == "pdh" {
+		logger.Info("Using performance data helper from PHD.dll for performance counter collection. This is in experimental state.")
+
+		toggle.PHDEnabled = true
+	}
+
 	if *printCollectors {
 		printCollectorsToStdout()
 
@@ -221,12 +227,6 @@ func run() int {
 
 	logger.Info("Enabled collectors: " + strings.Join(enabledCollectorList, ", "))
 
-	if v, ok := os.LookupEnv("WINDOWS_EXPORTER_PERF_COUNTERS_ENGINE"); ok && v == "pdh" || *togglePDH == "pdh" {
-		logger.Info("Using performance data helper from PHD.dll for performance counter collection. This is in experimental state.")
-
-		toggle.PHDEnabled = true
-	}
-
 	mux := http.NewServeMux()
 	mux.Handle("GET /health", httphandler.NewHealthHandler())
 	mux.Handle("GET /version", httphandler.NewVersionHandler())

internal/collector/cpu/cpu.go

@@ -109,7 +109,7 @@ func (c *Collector) Build(_ *slog.Logger, _ *mi.Session) error {
 
 		var err error
 
-		c.perfDataCollector, err = perfdata.NewCollector(perfdata.V1, "Processor Information", perfdata.AllInstances, counters)
+		c.perfDataCollector, err = perfdata.NewCollector(perfdata.V2, "Processor Information", perfdata.AllInstances, counters)
 		if err != nil {
 			return fmt.Errorf("failed to create Processor Information collector: %w", err)
 		}

internal/collector/system/const.go

@@ -0,0 +1,11 @@
+package system
+
+const (
+	ContextSwitchesPersec     = "Context Switches/sec"
+	ExceptionDispatchesPersec = "Exception Dispatches/sec"
+	ProcessorQueueLength      = "Processor Queue Length"
+	SystemCallsPersec         = "System Calls/sec"
+	SystemUpTime              = "System Up Time"
+	Processes                 = "Processes"
+	Threads                   = "Threads"
+)

internal/collector/system/system.go

@@ -4,11 +4,13 @@ package system
 
 import (
 	"errors"
+	"fmt"
 	"log/slog"
 
 	"github.com/alecthomas/kingpin/v2"
 	"github.com/prometheus-community/windows_exporter/internal/mi"
-	v1 "github.com/prometheus-community/windows_exporter/internal/perfdata/v1"
+	"github.com/prometheus-community/windows_exporter/internal/perfdata"
+	"github.com/prometheus-community/windows_exporter/internal/perfdata/perftypes"
 	"github.com/prometheus-community/windows_exporter/internal/types"
 	"github.com/prometheus/client_golang/prometheus"
 )
@@ -23,6 +25,8 @@ var ConfigDefaults = Config{}
 type Collector struct {
 	config Config
 
+	perfDataCollector perfdata.Collector
+
 	contextSwitchesTotal     *prometheus.Desc
 	exceptionDispatchesTotal *prometheus.Desc
 	processorQueueLength     *prometheus.Desc
@@ -54,14 +58,31 @@ func (c *Collector) GetName() string {
 }
 
 func (c *Collector) GetPerfCounter(_ *slog.Logger) ([]string, error) {
-	return []string{"System"}, nil
+	return []string{}, nil
 }
 
 func (c *Collector) Close(_ *slog.Logger) error {
 	return nil
 }
 
 func (c *Collector) Build(_ *slog.Logger, _ *mi.Session) error {
+	counters := []string{
+		ContextSwitchesPersec,
+		ExceptionDispatchesPersec,
+		ProcessorQueueLength,
+		SystemCallsPersec,
+		SystemUpTime,
+		Processes,
+		Threads,
+	}
+
+	var err error
+
+	c.perfDataCollector, err = perfdata.NewCollector(perfdata.V2, "System", nil, counters)
+	if err != nil {
+		return fmt.Errorf("failed to create System collector: %w", err)
+	}
+
 	c.contextSwitchesTotal = prometheus.NewDesc(
 		prometheus.BuildFQName(types.Namespace, Name, "context_switches_total"),
 		"Total number of context switches (WMI source: PerfOS_System.ContextSwitchesPersec)",
@@ -117,78 +138,59 @@ func (c *Collector) Build(_ *slog.Logger, _ *mi.Session) error {
 
 // Collect sends the metric values for each metric
 // to the provided prometheus Metric channel.
-func (c *Collector) Collect(ctx *types.ScrapeContext, logger *slog.Logger, ch chan<- prometheus.Metric) error {
-	logger = logger.With(slog.String("collector", Name))
-	if err := c.collect(ctx, logger, ch); err != nil {
-		logger.Error("failed collecting system metrics",
-			slog.Any("err", err),
-		)
-
-		return err
+func (c *Collector) Collect(_ *types.ScrapeContext, _ *slog.Logger, ch chan<- prometheus.Metric) error {
+	if err := c.collect(ch); err != nil {
+		return fmt.Errorf("failed collecting system metrics: %w", err)
 	}
 
 	return nil
 }
 
-// Win32_PerfRawData_PerfOS_System docs:
-// - https://web.archive.org/web/20050830140516/http://msdn.microsoft.com/library/en-us/wmisdk/wmi/win32_perfrawdata_perfos_system.asp
-type system struct {
-	ContextSwitchesPersec     float64 `perflib:"Context Switches/sec"`
-	ExceptionDispatchesPersec float64 `perflib:"Exception Dispatches/sec"`
-	ProcessorQueueLength      float64 `perflib:"Processor Queue Length"`
-	SystemCallsPersec         float64 `perflib:"System Calls/sec"`
-	SystemUpTime              float64 `perflib:"System Up Time"`
-	Processes                 float64 `perflib:"Processes"`
-	Threads                   float64 `perflib:"Threads"`
-}
-
-func (c *Collector) collect(ctx *types.ScrapeContext, logger *slog.Logger, ch chan<- prometheus.Metric) error {
-	logger = logger.With(slog.String("collector", Name))
-
-	var dst []system
-
-	if err := v1.UnmarshalObject(ctx.PerfObjects["System"], &dst, logger); err != nil {
-		return err
+func (c *Collector) collect(ch chan<- prometheus.Metric) error {
+	perfData, err := c.perfDataCollector.Collect()
+	if err != nil {
+		return fmt.Errorf("failed to collect System metrics: %w", err)
 	}
 
-	if len(dst) == 0 {
-		return errors.New("no data returned from Performance Counter")
+	data, ok := perfData[perftypes.EmptyInstance]
+	if !ok {
+		return errors.New("query for System returned empty result set")
 	}
 
 	ch <- prometheus.MustNewConstMetric(
 		c.contextSwitchesTotal,
 		prometheus.CounterValue,
-		dst[0].ContextSwitchesPersec,
+		data[ContextSwitchesPersec].FirstValue,
 	)
 	ch <- prometheus.MustNewConstMetric(
 		c.exceptionDispatchesTotal,
 		prometheus.CounterValue,
-		dst[0].ExceptionDispatchesPersec,
+		data[ExceptionDispatchesPersec].FirstValue,
 	)
 	ch <- prometheus.MustNewConstMetric(
 		c.processorQueueLength,
 		prometheus.GaugeValue,
-		dst[0].ProcessorQueueLength,
+		data[ProcessorQueueLength].FirstValue,
 	)
 	ch <- prometheus.MustNewConstMetric(
 		c.processes,
 		prometheus.GaugeValue,
-		dst[0].Processes,
+		data[Processes].FirstValue,
 	)
 	ch <- prometheus.MustNewConstMetric(
 		c.systemCallsTotal,
 		prometheus.CounterValue,
-		dst[0].SystemCallsPersec,
+		data[SystemCallsPersec].FirstValue,
 	)
 	ch <- prometheus.MustNewConstMetric(
 		c.systemUpTime,
 		prometheus.GaugeValue,
-		dst[0].SystemUpTime,
+		data[SystemUpTime].FirstValue,
 	)
 	ch <- prometheus.MustNewConstMetric(
 		c.threads,
 		prometheus.GaugeValue,
-		dst[0].Threads,
+		data[Threads].FirstValue,
 	)
 
 	// Windows has no defined limit, and is based off available resources. This currently isn't calculated by WMI and is set to default value.

internal/collector/terminal_services/const.go

@@ -1,23 +1,23 @@
 package terminal_services
 
 const (
-	HandleCount           = "Handle Count"
-	PageFaultsPersec      = "Page Faults/sec"
-	PageFileBytes         = "Page File Bytes"
-	PageFileBytesPeak     = "Page File Bytes Peak"
-	PercentPrivilegedTime = "% Privileged Time"
-	PercentProcessorTime  = "% Processor Time"
-	PercentUserTime       = "% User Time"
-	PoolNonpagedBytes     = "Pool Nonpaged Bytes"
-	PoolPagedBytes        = "Pool Paged Bytes"
-	PrivateBytes          = "Private Bytes"
-	ThreadCount           = "Thread Count"
-	VirtualBytes          = "Virtual Bytes"
-	VirtualBytesPeak      = "Virtual Bytes Peak"
-	WorkingSet            = "Working Set"
-	WorkingSetPeak        = "Working Set Peak"
+	handleCount           = "Handle Count"
+	pageFaultsPersec      = "Page Faults/sec"
+	pageFileBytes         = "Page File Bytes"
+	pageFileBytesPeak     = "Page File Bytes Peak"
+	percentPrivilegedTime = "% Privileged Time"
+	percentProcessorTime  = "% Processor Time"
+	percentUserTime       = "% User Time"
+	poolNonpagedBytes     = "Pool Nonpaged Bytes"
+	poolPagedBytes        = "Pool Paged Bytes"
+	privateBytes          = "Private Bytes"
+	threadCount           = "Thread Count"
+	virtualBytes          = "Virtual Bytes"
+	virtualBytesPeak      = "Virtual Bytes Peak"
+	workingSet            = "Working Set"
+	workingSetPeak        = "Working Set Peak"
 
-	SuccessfulConnections = "Successful Connections"
-	PendingConnections    = "Pending Connections"
-	FailedConnections     = "Failed Connections"
+	successfulConnections = "Successful Connections"
+	pendingConnections    = "Pending Connections"
+	failedConnections     = "Failed Connections"
 )