feat: Enhancing ETW logging for CNI (#4426)

Enahancing ETW logging for CNS

Author: apontejaj

cni/log/logger.go

@@ -38,7 +38,7 @@ func initZapLog(logFile string) *zap.Logger {
 		// If we fail to join the platform cores, fallback to the original core.
 		core = textFileCore
 	}
-	return zap.New(core, zap.AddCaller()).With(zap.Int("pid", os.Getpid()))
+	return zap.New(&metadataCore{Core: core}, zap.AddCaller()).With(zap.Int("pid", os.Getpid()))
 }
 
 var (

cni/log/metadata_core.go

@@ -0,0 +1,50 @@
+package log
+
+import (
+	"fmt"
+	"sync/atomic"
+
+	"go.uber.org/zap"
+	"go.uber.org/zap/zapcore"
+)
+
+// globalMetadata holds metadata fields shared across all loggers built from initZapLog.
+// It is set once from main() via SetMetadata after host metadata becomes available.
+var globalMetadata atomic.Pointer[[]zapcore.Field]
+
+// SetMetadata sets metadata fields that will be appended to every log entry
+// written through CNILogger, IPamLogger, and TelemetryLogger (and any loggers
+// derived from them). Safe to call once from main() after metadata is available.
+func SetMetadata(fields ...zap.Field) {
+	globalMetadata.Store(&fields)
+}
+
+// metadataCore is a zapcore.Core wrapper that appends shared metadata fields
+// to every log entry at write time. All cores derived via With() share the
+// same globalMetadata pointer, so setting metadata once propagates everywhere.
+type metadataCore struct {
+	zapcore.Core
+}
+
+func (c *metadataCore) With(fields []zapcore.Field) zapcore.Core {
+	return &metadataCore{Core: c.Core.With(fields)}
+}
+
+// Check registers this wrapper (not the inner core) as the writer so that
+// Write is called on metadataCore, allowing metadata fields to be appended.
+func (c *metadataCore) Check(entry zapcore.Entry, ce *zapcore.CheckedEntry) *zapcore.CheckedEntry {
+	if c.Enabled(entry.Level) {
+		return ce.AddCore(entry, c)
+	}
+	return ce
+}
+
+func (c *metadataCore) Write(entry zapcore.Entry, fields []zapcore.Field) error {
+	if extra := globalMetadata.Load(); extra != nil {
+		fields = append(fields, *extra...)
+	}
+	if err := c.Core.Write(entry, fields); err != nil {
+		return fmt.Errorf("writing log entry: %w", err)
+	}
+	return nil
+}

cni/network/plugin/main.go

@@ -6,6 +6,7 @@ package main
 import (
 	"fmt"
 	"os"
+	"path/filepath"
 	"time"
 
 	"github.com/Azure/azure-container-networking/cni"
@@ -50,6 +51,25 @@ func printVersion() {
 }
 
 func rootExecute() error {
+	// Enrich all CNI loggers with host metadata so ETW events carry VM identity for diagnostics.
+	metadataFile := filepath.Join(os.TempDir(), "azuremetadata.json")
+	if metadata, err := common.GetHostMetadata(metadataFile); err == nil {
+		host, _ := os.Hostname()
+		zaplog.SetMetadata(
+			zap.String("hostname", host),
+			zap.String("version", version),
+			zap.String("kubernetes_apiserver", os.Getenv("KUBERNETES_SERVICE_HOST")),
+			zap.String("account", metadata.SubscriptionID),
+			zap.String("anonymous_user_id", metadata.VMName),
+			zap.String("location", metadata.Location),
+			zap.String("resource_group", metadata.ResourceGroupName),
+			zap.String("vm_size", metadata.VMSize),
+			zap.String("os_version", metadata.OSVersion),
+			zap.String("vm_id", metadata.VMID),
+			zap.String("session_id", metadata.VMID),
+		)
+	}
+
 	var config common.PluginConfig
 
 	config.Version = version

cni/network/stateless/main.go

@@ -6,6 +6,7 @@ package main
 import (
 	"fmt"
 	"os"
+	"path/filepath"
 	"time"
 
 	"github.com/Azure/azure-container-networking/cni"
@@ -50,6 +51,25 @@ func printVersion() {
 }
 
 func rootExecute() error {
+	// Enrich all CNI loggers with host metadata so ETW events carry VM identity for diagnostics.
+	metadataFile := filepath.Join(os.TempDir(), "azuremetadata.json")
+	if metadata, err := common.GetHostMetadata(metadataFile); err == nil {
+		host, _ := os.Hostname()
+		zapLog.SetMetadata(
+			zap.String("hostname", host),
+			zap.String("version", version),
+			zap.String("kubernetes_apiserver", os.Getenv("KUBERNETES_SERVICE_HOST")),
+			zap.String("account", metadata.SubscriptionID),
+			zap.String("anonymous_user_id", metadata.VMName),
+			zap.String("location", metadata.Location),
+			zap.String("resource_group", metadata.ResourceGroupName),
+			zap.String("vm_size", metadata.VMSize),
+			zap.String("os_version", metadata.OSVersion),
+			zap.String("vm_id", metadata.VMID),
+			zap.String("session_id", metadata.VMID),
+		)
+	}
+
 	var config common.PluginConfig
 
 	log.SetName(name)