Add NVIDIA Triton GMP integration (#881)

Adds a NVIDIA Triton sample dashboard, associated with integration metadata (currently HIDDEN; will launch after validation and docs updates). go/triton-dashboard-dev-notes (internal link)
GoogleCloudPlatform · Dec 18, 2024 · bd8167e · bd8167e
1 parent 5437a1b
commit bd8167e
Show file tree

Hide file tree

Showing 7 changed files with 1,212 additions and 0 deletions.
diff --git a/dashboards/nvidia-gpu/README.md b/dashboards/nvidia-gpu/README.md
@@ -21,3 +21,8 @@
 |:------------------|
 |Filename: [nvidia-dcgm-prometheus.json](nvidia-dcgm-prometheus.json)|
 |Displays Advanced GPU metrics from NVIDIA Datacenter GPU Manager (DCGM).  This requires a specific setup (e.g. installing DCGM and DCGM exporter).|
+
+|NVIDIA Triton Inference Server (GKE Only)|
+|:------------------|
+|Filename: [nvidia-triton-prometheus.json](nvidia-triton-prometheus.json)|
+|Displays Triton Inference Server metrics. This requires your GKE cluster to have NVIDIA Triton installed and exporting metrics via Managed Prometheus.|
diff --git a/dashboards/nvidia-gpu/metadata.yaml b/dashboards/nvidia-gpu/metadata.yaml
@@ -23,3 +23,12 @@ sample_dashboards:
     related_integrations:
       - id: dcgm
         platform: GKE
+  -
+    category: NVIDIA GPUs
+    id: nvidia-triton-prometheus
+    display_name: NVIDIA Triton Inference Server (GKE Only)
+    description: |-
+      Displays Triton Inference Server metrics. This requires your GKE cluster to have NVIDIA Triton installed and exporting metrics via Managed Prometheus.
+    related_integrations:
+      - id: nvidia-triton
+        platform: GKE
diff --git a/dashboards/nvidia-gpu/nvidia-triton-prometheus.01.png b/dashboards/nvidia-gpu/nvidia-triton-prometheus.01.png
diff --git a/dashboards/nvidia-gpu/nvidia-triton-prometheus.02.png b/dashboards/nvidia-gpu/nvidia-triton-prometheus.02.png