enhance: [2.5] Add metrics for parse expression (#39716)

master pr: #39654

Signed-off-by: Cai Zhang <cai.zhang@zilliz.com>

internal/proxy/task_delete.go

@@ -4,6 +4,8 @@ import (
 	"context"
 	"fmt"
 	"io"
+	"strconv"
+	"time"
 
 	"github.com/cockroachdb/errors"
 	"go.opentelemetry.io/otel"
@@ -347,10 +349,13 @@ func (dr *deleteRunner) Init(ctx context.Context) error {
 		return ErrWithLog(log, "Failed to get collection schema", err)
 	}
 
+	start := time.Now()
 	dr.plan, err = planparserv2.CreateRetrievePlan(dr.schema.schemaHelper, dr.req.GetExpr(), dr.req.GetExprTemplateValues())
 	if err != nil {
+		metrics.ProxyParseExpressionLatency.WithLabelValues(strconv.FormatInt(paramtable.GetNodeID(), 10), "delete", metrics.FailLabel).Observe(float64(time.Since(start).Milliseconds()))
 		return merr.WrapErrAsInputError(merr.WrapErrParameterInvalidMsg("failed to create delete plan: %v", err))
 	}
+	metrics.ProxyParseExpressionLatency.WithLabelValues(strconv.FormatInt(paramtable.GetNodeID(), 10), "delete", metrics.SuccessLabel).Observe(float64(time.Since(start).Milliseconds()))
 
 	if planparserv2.IsAlwaysTruePlan(dr.plan) {
 		return merr.WrapErrAsInputError(merr.WrapErrParameterInvalidMsg("delete plan can't be empty or always true : %s", dr.req.GetExpr()))

internal/proxy/task_query.go

@@ -5,6 +5,7 @@ import (
 	"fmt"
 	"strconv"
 	"strings"
+	"time"
 
 	"github.com/cockroachdb/errors"
 	"github.com/samber/lo"
@@ -235,11 +236,13 @@ func createCntPlan(expr string, schemaHelper *typeutil.SchemaHelper, exprTemplat
 			},
 		}, nil
 	}
+	start := time.Now()
 	plan, err := planparserv2.CreateRetrievePlan(schemaHelper, expr, exprTemplateValues)
 	if err != nil {
+		metrics.ProxyParseExpressionLatency.WithLabelValues(strconv.FormatInt(paramtable.GetNodeID(), 10), "query", metrics.FailLabel).Observe(float64(time.Since(start).Milliseconds()))
 		return nil, merr.WrapErrAsInputError(merr.WrapErrParameterInvalidMsg("failed to create query plan: %v", err))
 	}
-
+	metrics.ProxyParseExpressionLatency.WithLabelValues(strconv.FormatInt(paramtable.GetNodeID(), 10), "query", metrics.SuccessLabel).Observe(float64(time.Since(start).Milliseconds()))
 	plan.Node.(*planpb.PlanNode_Query).Query.IsCount = true
 
 	return plan, nil
@@ -258,10 +261,13 @@ func (t *queryTask) createPlan(ctx context.Context) error {
 
 	var err error
 	if t.plan == nil {
+		start := time.Now()
 		t.plan, err = planparserv2.CreateRetrievePlan(schema.schemaHelper, t.request.Expr, t.request.GetExprTemplateValues())
 		if err != nil {
+			metrics.ProxyParseExpressionLatency.WithLabelValues(strconv.FormatInt(paramtable.GetNodeID(), 10), "query", metrics.FailLabel).Observe(float64(time.Since(start).Milliseconds()))
 			return merr.WrapErrAsInputError(merr.WrapErrParameterInvalidMsg("failed to create query plan: %v", err))
 		}
+		metrics.ProxyParseExpressionLatency.WithLabelValues(strconv.FormatInt(paramtable.GetNodeID(), 10), "query", metrics.SuccessLabel).Observe(float64(time.Since(start).Milliseconds()))
 	}
 
 	t.request.OutputFields, t.userOutputFields, t.userDynamicFields, err = translateOutputFields(t.request.OutputFields, t.schema, true)

internal/proxy/task_search.go

@@ -5,6 +5,7 @@ import (
 	"fmt"
 	"math"
 	"strconv"
+	"time"
 
 	"github.com/cockroachdb/errors"
 	"github.com/samber/lo"
@@ -534,13 +535,16 @@ func (t *searchTask) tryGeneratePlan(params []*commonpb.KeyValuePair, dsl string
 	}
 
 	searchInfo.planInfo.QueryFieldId = annField.GetFieldID()
+	start := time.Now()
 	plan, planErr := planparserv2.CreateSearchPlan(t.schema.schemaHelper, dsl, annsFieldName, searchInfo.planInfo, exprTemplateValues)
 	if planErr != nil {
 		log.Ctx(t.ctx).Warn("failed to create query plan", zap.Error(planErr),
 			zap.String("dsl", dsl), // may be very large if large term passed.
 			zap.String("anns field", annsFieldName), zap.Any("query info", searchInfo.planInfo))
+		metrics.ProxyParseExpressionLatency.WithLabelValues(strconv.FormatInt(paramtable.GetNodeID(), 10), "search", metrics.FailLabel).Observe(float64(time.Since(start).Milliseconds()))
 		return nil, nil, 0, false, merr.WrapErrParameterInvalidMsg("failed to create query plan: %v", planErr)
 	}
+	metrics.ProxyParseExpressionLatency.WithLabelValues(strconv.FormatInt(paramtable.GetNodeID(), 10), "search", metrics.SuccessLabel).Observe(float64(time.Since(start).Milliseconds()))
 	log.Ctx(t.ctx).Debug("create query plan",
 		zap.String("dsl", t.request.Dsl), // may be very large if large term passed.
 		zap.String("anns field", annsFieldName), zap.Any("query info", searchInfo.planInfo))

pkg/metrics/proxy_metrics.go

@@ -427,6 +427,15 @@ var (
 			Help:      "the number of non-zeros in each sparse search task",
 			Buckets:   buckets,
 		}, []string{nodeIDLabelName, collectionName})
+
+	ProxyParseExpressionLatency = prometheus.NewHistogramVec(
+		prometheus.HistogramOpts{
+			Namespace: milvusNamespace,
+			Subsystem: typeutil.ProxyRole,
+			Name:      "parse_expr_latency",
+			Help:      "the latency of parse expression",
+			Buckets:   buckets,
+		}, []string{nodeIDLabelName, functionLabelName, statusLabelName})
 )
 
 // RegisterProxy registers Proxy metrics
@@ -491,6 +500,8 @@ func RegisterProxy(registry *prometheus.Registry) {
 
 	registry.MustRegister(ProxySearchSparseNumNonZeros)
 
+	registry.MustRegister(ProxyParseExpressionLatency)
+
 	RegisterStreamingServiceClient(registry)
 }