From 9161c04ddb46e885e1bf71ef2ca4410514d96c2a Mon Sep 17 00:00:00 2001
From: Zachary Bennett <bennett.zachary@outlook.com>
Date: Tue, 24 Feb 2026 12:32:11 -0600
Subject: [PATCH] docs: add histogram latency metrics documentation and test
 patterns

Update metrics documentation and test patterns to include the 5 new
histogram latency metrics added in triton-inference-server/core.

- Document all 6 histogram metrics in the metrics user guide table
- List all histogram families that support custom bucket boundaries
- Update INF_HISTOGRAM_PATTERNS in metrics_config_test.py

Companion to triton-inference-server/core histogram latency metrics PR.

Addresses #7672
---
 docs/user_guide/metrics.md           | 14 ++++++++++++--
 qa/L0_metrics/metrics_config_test.py | 11 +++++++++--
 2 files changed, 21 insertions(+), 4 deletions(-)

diff --git a/docs/user_guide/metrics.md b/docs/user_guide/metrics.md
index 3d41872f42..7458622fb0 100644
--- a/docs/user_guide/metrics.md
+++ b/docs/user_guide/metrics.md
@@ -1,5 +1,5 @@
 <!--
-# Copyright 2018-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright 2018-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 #
 # Redistribution and use in source and binary forms, with or without
 # modification, are permitted provided that the following conditions
@@ -218,6 +218,11 @@ metrics are used for latencies:
 |Category      |Metric          |Metric Name |Description                |Granularity|Frequency    |
 |--------------|----------------|------------|---------------------------|-----------|-------------|
 |Latency       |Request to First Response Time    |`nv_inference_first_response_histogram_ms` |Histogram of end-to-end inference request to the first response time |Per model  |Per request  |
+|              |Request Time    |`nv_inference_request_duration_histogram_us` |Histogram of end-to-end inference request duration (includes cached requests) |Per model  |Per request  |
+|              |Queue Time      |`nv_inference_queue_duration_histogram_us` |Histogram of time requests spend waiting in the scheduling queue |Per model  |Per request  |
+|              |Compute Input Time|`nv_inference_compute_input_duration_histogram_us` |Histogram of time requests spend processing inference inputs |Per model  |Per request  |
+|              |Compute Time    |`nv_inference_compute_infer_duration_histogram_us` |Histogram of time requests spend executing the inference model |Per model  |Per request  |
+|              |Compute Output Time|`nv_inference_compute_output_duration_histogram_us` |Histogram of time requests spend processing inference outputs |Per model  |Per request  |
 
 To enable these metrics specifically, you can set `--metrics-config histogram_latencies=true`
 
@@ -265,7 +270,12 @@ model_metrics {
 
 Currently, the following histogram families support custom buckets.
 ```
-nv_inference_first_response_histogram_ms  // Time to First Response
+nv_inference_first_response_histogram_ms           // Time to First Response (ms)
+nv_inference_request_duration_histogram_us          // Request Duration (us)
+nv_inference_queue_duration_histogram_us            // Queue Duration (us)
+nv_inference_compute_input_duration_histogram_us    // Compute Input Duration (us)
+nv_inference_compute_infer_duration_histogram_us    // Compute Infer Duration (us)
+nv_inference_compute_output_duration_histogram_us   // Compute Output Duration (us)
 ```
 
 #### Summaries
diff --git a/qa/L0_metrics/metrics_config_test.py b/qa/L0_metrics/metrics_config_test.py
index 0c448dbaa7..7514ef2ff1 100755
--- a/qa/L0_metrics/metrics_config_test.py
+++ b/qa/L0_metrics/metrics_config_test.py
@@ -1,5 +1,5 @@
 #!/usr/bin/python
-# Copyright 2023-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright 2023-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 #
 # Redistribution and use in source and binary forms, with or without
 # modification, are permitted provided that the following conditions
@@ -44,7 +44,14 @@
     "nv_inference_compute_infer_duration",
     "nv_inference_compute_output_duration",
 ]
-INF_HISTOGRAM_PATTERNS = ["nv_inference_first_response_histogram_ms"]
+INF_HISTOGRAM_PATTERNS = [
+    "nv_inference_first_response_histogram_ms",
+    "nv_inference_request_duration_histogram_us",
+    "nv_inference_queue_duration_histogram_us",
+    "nv_inference_compute_input_duration_histogram_us",
+    "nv_inference_compute_infer_duration_histogram_us",
+    "nv_inference_compute_output_duration_histogram_us",
+]
 INF_SUMMARY_PATTERNS = [
     "nv_inference_request_summary",
     "nv_inference_queue_summary",