fix subsystem prefix in processing latency metric tests

satyamg1620 · satyamg1620 · commit 8bde1d283664 · 2026-07-20T18:37:31.000+05:30
Signed-off-by: satyamg1620 &lt;Satyam.Gupta.3@ibm.com&gt;
diff --git a/docs/metrics.md b/docs/metrics.md
@@ -103,6 +103,15 @@ Label `{name}` (the pool name).
 | `scheduler_e2e_duration_seconds` | Histogram | End-to-end scheduling latency. |
 | `scheduler_attempts_total` | Counter | Scheduling attempts; labels `{status, target_model_name, endpoint_name, namespace, port}`. |
 
+### EPP processing overhead
+
+Unlabelled. Together these cover the EPP's own cost on the request and response paths.
+
+| Name | Type | Notes |
+|---|---|---|
+| `request_processing_duration_seconds` | Histogram | Request orchestration latency, from receipt through endpoint selection and request preparation; excludes admission control. |
+| `response_processing_duration_seconds` | Histogram | Response handling latency, accumulated across the response handlers; excludes model-server generation time. |
+
 ### Plugin, info, and model rewrite
 
 | Name | Type | Notes |
diff --git a/pkg/epp/metrics/metrics.go b/pkg/epp/metrics/metrics.go
@@ -756,7 +756,8 @@ func RecordSchedulerE2ELatency(duration time.Duration) {
 }
 
 // RecordRequestProcessingLatency records the EPP request processing latency,
-// measured from request receipt until an endpoint is selected.
+// measured from request receipt through endpoint selection and request
+// preparation, excluding admission-control time.
 func RecordRequestProcessingLatency(duration time.Duration) {
 	llmdRequestProcessingLatency.WithLabelValues().Observe(duration.Seconds())
 }
diff --git a/pkg/epp/metrics/metrics_test.go b/pkg/epp/metrics/metrics_test.go
@@ -903,7 +903,7 @@ func TestRequestProcessingLatency(t *testing.T) {
 		t.Fatal(err)
 	}
 	defer want.Close()
-	if err := promtestutil.GatherAndCompare(metrics.Registry, want, "llm_d_router_epp_request_processing_duration_seconds"); err != nil {
+	if err := promtestutil.GatherAndCompare(metrics.Registry, want, "llm_d_epp_request_processing_duration_seconds"); err != nil {
 		t.Error(err)
 	}
 }
@@ -930,7 +930,7 @@ func TestResponseProcessingLatency(t *testing.T) {
 		t.Fatal(err)
 	}
 	defer want.Close()
-	if err := promtestutil.GatherAndCompare(metrics.Registry, want, "llm_d_router_epp_response_processing_duration_seconds"); err != nil {
+	if err := promtestutil.GatherAndCompare(metrics.Registry, want, "llm_d_epp_response_processing_duration_seconds"); err != nil {
 		t.Error(err)
 	}
 }
diff --git a/pkg/epp/metrics/testdata/llm_d_request_processing_duration_seconds_metric b/pkg/epp/metrics/testdata/llm_d_request_processing_duration_seconds_metric
@@ -1,15 +1,15 @@
-# HELP llm_d_router_epp_request_processing_duration_seconds [ALPHA] EPP request orchestration latency distribution in seconds, from request receipt through endpoint selection and request preparation, excluding admission-control time.
-# TYPE llm_d_router_epp_request_processing_duration_seconds histogram
-llm_d_router_epp_request_processing_duration_seconds_bucket{le="0.0001"} 0
-llm_d_router_epp_request_processing_duration_seconds_bucket{le="0.0002"} 1
-llm_d_router_epp_request_processing_duration_seconds_bucket{le="0.0005"} 1
-llm_d_router_epp_request_processing_duration_seconds_bucket{le="0.001"} 2
-llm_d_router_epp_request_processing_duration_seconds_bucket{le="0.002"} 3
-llm_d_router_epp_request_processing_duration_seconds_bucket{le="0.005"} 4
-llm_d_router_epp_request_processing_duration_seconds_bucket{le="0.01"} 5
-llm_d_router_epp_request_processing_duration_seconds_bucket{le="0.02"} 6
-llm_d_router_epp_request_processing_duration_seconds_bucket{le="0.05"} 7
-llm_d_router_epp_request_processing_duration_seconds_bucket{le="0.1"} 8
-llm_d_router_epp_request_processing_duration_seconds_bucket{le="+Inf"} 9
-llm_d_router_epp_request_processing_duration_seconds_sum 0.2835
-llm_d_router_epp_request_processing_duration_seconds_count 9
+# HELP llm_d_epp_request_processing_duration_seconds [ALPHA] EPP request orchestration latency distribution in seconds, from request receipt through endpoint selection and request preparation, excluding admission-control time.
+# TYPE llm_d_epp_request_processing_duration_seconds histogram
+llm_d_epp_request_processing_duration_seconds_bucket{le="0.0001"} 0
+llm_d_epp_request_processing_duration_seconds_bucket{le="0.0002"} 1
+llm_d_epp_request_processing_duration_seconds_bucket{le="0.0005"} 1
+llm_d_epp_request_processing_duration_seconds_bucket{le="0.001"} 2
+llm_d_epp_request_processing_duration_seconds_bucket{le="0.002"} 3
+llm_d_epp_request_processing_duration_seconds_bucket{le="0.005"} 4
+llm_d_epp_request_processing_duration_seconds_bucket{le="0.01"} 5
+llm_d_epp_request_processing_duration_seconds_bucket{le="0.02"} 6
+llm_d_epp_request_processing_duration_seconds_bucket{le="0.05"} 7
+llm_d_epp_request_processing_duration_seconds_bucket{le="0.1"} 8
+llm_d_epp_request_processing_duration_seconds_bucket{le="+Inf"} 9
+llm_d_epp_request_processing_duration_seconds_sum 0.2835
+llm_d_epp_request_processing_duration_seconds_count 9
diff --git a/pkg/epp/metrics/testdata/llm_d_response_processing_duration_seconds_metric b/pkg/epp/metrics/testdata/llm_d_response_processing_duration_seconds_metric
@@ -1,15 +1,15 @@
-# HELP llm_d_router_epp_response_processing_duration_seconds [ALPHA] EPP response processing latency distribution in seconds, accumulated across response handlers and excluding model server generation time.
-# TYPE llm_d_router_epp_response_processing_duration_seconds histogram
-llm_d_router_epp_response_processing_duration_seconds_bucket{le="0.0001"} 0
-llm_d_router_epp_response_processing_duration_seconds_bucket{le="0.0002"} 1
-llm_d_router_epp_response_processing_duration_seconds_bucket{le="0.0005"} 1
-llm_d_router_epp_response_processing_duration_seconds_bucket{le="0.001"} 2
-llm_d_router_epp_response_processing_duration_seconds_bucket{le="0.002"} 3
-llm_d_router_epp_response_processing_duration_seconds_bucket{le="0.005"} 4
-llm_d_router_epp_response_processing_duration_seconds_bucket{le="0.01"} 5
-llm_d_router_epp_response_processing_duration_seconds_bucket{le="0.02"} 6
-llm_d_router_epp_response_processing_duration_seconds_bucket{le="0.05"} 7
-llm_d_router_epp_response_processing_duration_seconds_bucket{le="0.1"} 8
-llm_d_router_epp_response_processing_duration_seconds_bucket{le="+Inf"} 9
-llm_d_router_epp_response_processing_duration_seconds_sum 0.2835
-llm_d_router_epp_response_processing_duration_seconds_count 9
+# HELP llm_d_epp_response_processing_duration_seconds [ALPHA] EPP response processing latency distribution in seconds, accumulated across response handlers and excluding model server generation time.
+# TYPE llm_d_epp_response_processing_duration_seconds histogram
+llm_d_epp_response_processing_duration_seconds_bucket{le="0.0001"} 0
+llm_d_epp_response_processing_duration_seconds_bucket{le="0.0002"} 1
+llm_d_epp_response_processing_duration_seconds_bucket{le="0.0005"} 1
+llm_d_epp_response_processing_duration_seconds_bucket{le="0.001"} 2
+llm_d_epp_response_processing_duration_seconds_bucket{le="0.002"} 3
+llm_d_epp_response_processing_duration_seconds_bucket{le="0.005"} 4
+llm_d_epp_response_processing_duration_seconds_bucket{le="0.01"} 5
+llm_d_epp_response_processing_duration_seconds_bucket{le="0.02"} 6
+llm_d_epp_response_processing_duration_seconds_bucket{le="0.05"} 7
+llm_d_epp_response_processing_duration_seconds_bucket{le="0.1"} 8
+llm_d_epp_response_processing_duration_seconds_bucket{le="+Inf"} 9
+llm_d_epp_response_processing_duration_seconds_sum 0.2835
+llm_d_epp_response_processing_duration_seconds_count 9

Original file line number	Diff line number	Diff line change
`@@ -756,7 +756,8 @@ func RecordSchedulerE2ELatency(duration time.Duration) {`
`756`	`756`	`}`
`757`	`757`
`758`	`758`	`// RecordRequestProcessingLatency records the EPP request processing latency,`
`759`		`-// measured from request receipt until an endpoint is selected.`
	`759`	`+// measured from request receipt through endpoint selection and request`
	`760`	`+// preparation, excluding admission-control time.`
`760`	`761`	`func RecordRequestProcessingLatency(duration time.Duration) {`
`761`	`762`	`llmdRequestProcessingLatency.WithLabelValues().Observe(duration.Seconds())`
`762`	`763`	`}`
Original file line number	Diff line number	Diff line change
`@@ -903,7 +903,7 @@ func TestRequestProcessingLatency(t *testing.T) {`
`903`	`903`	`t.Fatal(err)`
`904`	`904`	`}`
`905`	`905`	`defer want.Close()`
`906`		`- if err := promtestutil.GatherAndCompare(metrics.Registry, want, "llm_d_router_epp_request_processing_duration_seconds"); err != nil {`
	`906`	`+ if err := promtestutil.GatherAndCompare(metrics.Registry, want, "llm_d_epp_request_processing_duration_seconds"); err != nil {`
`907`	`907`	`t.Error(err)`
`908`	`908`	`}`
`909`	`909`	`}`
`@@ -930,7 +930,7 @@ func TestResponseProcessingLatency(t *testing.T) {`
`930`	`930`	`t.Fatal(err)`
`931`	`931`	`}`
`932`	`932`	`defer want.Close()`
`933`		`- if err := promtestutil.GatherAndCompare(metrics.Registry, want, "llm_d_router_epp_response_processing_duration_seconds"); err != nil {`
	`933`	`+ if err := promtestutil.GatherAndCompare(metrics.Registry, want, "llm_d_epp_response_processing_duration_seconds"); err != nil {`
`934`	`934`	`t.Error(err)`
`935`	`935`	`}`
`936`	`936`	`}`