Revert "Revert "wip: error code monitoring/metric recording""

This reverts commit 44c31be.

charts/dapr/crds/configuration.yaml

@@ -266,6 +266,8 @@ spec:
                         description: If true (default is false) HTTP verbs (e.g., GET, POST) are excluded from the metrics.
                         type: boolean
                     type: object
+                  recordErrorCodes:
+                    type: boolean
                   rules:
                     items:
                       description: MetricsRule defines configuration options for a
@@ -329,6 +331,8 @@ spec:
                     items:
                       type: integer
                     type: array
+                  recordErrorCodes:
+                    type: boolean
                   rules:
                     items:
                       description: MetricsRule defines configuration options for a

pkg/api/http/http.go

@@ -508,6 +508,7 @@ func (a *api) onBulkGetState(reqCtx *fasthttp.RequestCtx) {
 			status := apierrors.StateStore(storeName).InvalidKeyName(k, err.Error())
 			universalFastHTTPErrorResponder(reqCtx, status)
 			log.Debug(status)
+			diag.DefaultErrorCodeMonitoring.RecordErrorCode(errorcodes.IllegalKey)
 			return
 		}
 		r := state.GetRequest{

pkg/apis/configuration/v1alpha1/types.go

@@ -212,6 +212,8 @@ type ZipkinSpec struct {
 type MetricSpec struct {
 	Enabled *bool `json:"enabled"`
 	// +optional
+	RecordErrorCodes *bool `json:"recordErrorCodes,omitempty"`
+	// +optional
 	HTTP *MetricHTTP `json:"http,omitempty"`
 	// +optional
 	Rules []MetricsRule `json:"rules,omitempty"`

pkg/config/configuration.go

@@ -254,8 +254,9 @@ func (o OtelSpec) GetIsSecure() bool {
 // MetricSpec configuration for metrics.
 type MetricSpec struct {
 	// Defaults to true
-	Enabled *bool       `json:"enabled,omitempty" yaml:"enabled,omitempty"`
-	HTTP    *MetricHTTP `json:"http,omitempty" yaml:"http,omitempty"`
+	Enabled          *bool       `json:"enabled,omitempty" yaml:"enabled,omitempty"`
+	RecordErrorCodes *bool       `json:"recordErrorCodes,omitempty"`
+	HTTP             *MetricHTTP `json:"http,omitempty" yaml:"http,omitempty"`
 	// Latency distribution buckets. If not set, the default buckets are used.
 	LatencyDistributionBuckets *[]int        `json:"latencyDistributionBuckets,omitempty" yaml:"latencyDistributionBuckets,omitempty"`
 	Rules                      []MetricsRule `json:"rules,omitempty" yaml:"rules,omitempty"`
@@ -317,6 +318,15 @@ func (m MetricSpec) GetHTTPPathMatching() []string {
 	return m.HTTP.PathMatching
 }
 
+// GetRecordErrorCodes returns true if `recordErrorCodes` is enabled for metrics
+func (m MetricSpec) GetRecordErrorCodes() bool {
+	if m.RecordErrorCodes == nil {
+		// The default is false
+		return false
+	}
+	return *m.RecordErrorCodes
+}
+
 // MetricHTTP defines configuration for metrics for the HTTP server
 type MetricHTTP struct {
 	// If false, metrics for the HTTP server are collected with increased cardinality.

pkg/diagnostics/errorcode_monitoring.go

@@ -0,0 +1,51 @@
+package diagnostics
+
+import (
+	"context"
+
+	"go.opencensus.io/stats"
+	"go.opencensus.io/stats/view"
+	"go.opencensus.io/tag"
+
+	diagUtils "github.com/dapr/dapr/pkg/diagnostics/utils"
+)
+
+type errorCodeMetrics struct {
+	errorCodeCount *stats.Int64Measure
+
+	appID   string
+	ctx     context.Context
+	enabled bool
+}
+
+func newErrorCodeMetrics() *errorCodeMetrics {
+	return &errorCodeMetrics{ //nolint:exhaustruct
+		errorCodeCount: stats.Int64(
+			"error_code/count",
+			"Number of times an error with a specific errorcode was encountered.",
+			stats.UnitDimensionless),
+
+		ctx:     context.Background(),
+		enabled: false,
+	}
+}
+
+// Init registers the errorcode metrics view.
+func (m *errorCodeMetrics) Init(id string) error {
+	m.enabled = true
+	m.appID = id
+
+	return view.Register(
+		diagUtils.NewMeasureView(m.errorCodeCount, []tag.Key{appIDKey, errorCodeKey}, view.Count()),
+	)
+}
+
+func (m *errorCodeMetrics) RecordErrorCode(code string) {
+	if m.enabled {
+		_ = stats.RecordWithTags(
+			m.ctx,
+			diagUtils.WithTags(m.errorCodeCount.Name(), appIDKey, m.appID, errorCodeKey, code),
+			m.errorCodeCount.M(1),
+		)
+	}
+}

pkg/diagnostics/metrics.go

@@ -21,6 +21,7 @@ import (
 
 	"github.com/dapr/dapr/pkg/config"
 	"github.com/dapr/dapr/pkg/diagnostics/utils"
+	"github.com/dapr/dapr/pkg/messages/errorcodes"
 )
 
 // appIDKey is a tag key for App ID.
@@ -42,6 +43,8 @@ var (
 	DefaultResiliencyMonitoring = newResiliencyMetrics()
 	// DefaultWorkflowMonitoring holds workflow specific metrics.
 	DefaultWorkflowMonitoring = newWorkflowMetrics()
+	// DefaultErrorCodeMonitoring holds error code specific metrics.
+	DefaultErrorCodeMonitoring = newErrorCodeMetrics()
 )
 
 // <<10 -> KBs; <<20 -> MBs; <<30 -> GBs
@@ -79,6 +82,17 @@ func InitMetrics(appID, namespace string, metricSpec config.MetricSpec) error {
 		return err
 	}
 
+	log.Info("jake::: my build!!!!!!!!!!!!!")
+	if metricSpec.GetRecordErrorCodes() {
+		if err := DefaultErrorCodeMonitoring.Init(appID); err != nil {
+			return err
+		}
+		log.Info("jake::: error code monitoring success")
+		DefaultErrorCodeMonitoring.RecordErrorCode(errorcodes.ActorInstanceMissing)
+		DefaultErrorCodeMonitoring.RecordErrorCode(errorcodes.ActorInstanceMissing)
+		DefaultErrorCodeMonitoring.RecordErrorCode(errorcodes.PubsubEmpty)
+	}
+
 	// Set reporting period of views
 	view.SetReportingPeriod(DefaultReportingPeriod)
 	return utils.CreateRulesMap(metricSpec.Rules)

pkg/diagnostics/service_monitoring.go

@@ -23,6 +23,7 @@ var (
 	namespaceKey        = tag.MustNewKey("namespace")
 	resiliencyNameKey   = tag.MustNewKey("name")
 	policyKey           = tag.MustNewKey("policy")
+	errorCodeKey        = tag.MustNewKey("errorcode")
 	componentNameKey    = tag.MustNewKey("componentName")
 	destinationAppIDKey = tag.MustNewKey("dst_app_id")
 	sourceAppIDKey      = tag.MustNewKey("src_app_id")