From 5db6ec3e1fc8b2d30d3aa017f0ff7bc4dd943000 Mon Sep 17 00:00:00 2001
From: Arsolitt <arsolitt@gmail.com>
Date: Sat, 18 Apr 2026 07:27:33 +0300
Subject: [PATCH] fix(dashboard): scope GPU panels to selected namespace

Pod-level panels on the efficiency dashboard and DCGM-level panels on
the performance dashboard ignored the $namespace template variable, so
changing it left the visualizations unchanged. Add the filter to each
query. Performance-side queries use the `$namespace|` empty-tolerant
form so host-level DCGM series without a namespace label remain
visible when a specific namespace is selected.

Signed-off-by: Arsolitt <arsolitt@gmail.com>
---
 dashboards/gpu/gpu-efficiency.json  |  4 ++--
 dashboards/gpu/gpu-performance.json | 12 ++++++------
 2 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/dashboards/gpu/gpu-efficiency.json b/dashboards/gpu/gpu-efficiency.json
index dddb2001..e2e8c105 100644
--- a/dashboards/gpu/gpu-efficiency.json
+++ b/dashboards/gpu/gpu-efficiency.json
@@ -363,7 +363,7 @@
       "id": 21,
       "targets": [
         {
-          "expr": "topk(20, pod:tensor_saturation:avg5m * 100)",
+          "expr": "topk(20, pod:tensor_saturation:avg5m{namespace=~\"$namespace\"} * 100)",
           "instant": true,
           "range": false,
           "format": "table",
@@ -501,7 +501,7 @@
       "id": 22,
       "targets": [
         {
-          "expr": "topk(20, pod:util_per_watt:avg5m)",
+          "expr": "topk(20, pod:util_per_watt:avg5m{namespace=~\"$namespace\"})",
           "instant": true,
           "range": false,
           "format": "table",
diff --git a/dashboards/gpu/gpu-performance.json b/dashboards/gpu/gpu-performance.json
index b0f26958..e16f8c47 100644
--- a/dashboards/gpu/gpu-performance.json
+++ b/dashboards/gpu/gpu-performance.json
@@ -553,7 +553,7 @@
       "id": 22,
       "targets": [
         {
-          "expr": "DCGM_FI_DEV_FB_FREE{Hostname=~\"$Hostname\"} * 1048576",
+          "expr": "DCGM_FI_DEV_FB_FREE{Hostname=~\"$Hostname\", namespace=~\"$namespace|\"} * 1048576",
           "legendFormat": "{{Hostname}}/{{gpu}}",
           "refId": "A"
         }
@@ -617,7 +617,7 @@
       "id": 31,
       "targets": [
         {
-          "expr": "DCGM_FI_DEV_POWER_USAGE{Hostname=~\"$Hostname\"}",
+          "expr": "DCGM_FI_DEV_POWER_USAGE{Hostname=~\"$Hostname\", namespace=~\"$namespace|\"}",
           "legendFormat": "{{Hostname}}/{{gpu}}",
           "refId": "A"
         }
@@ -668,7 +668,7 @@
       "id": 32,
       "targets": [
         {
-          "expr": "DCGM_FI_DEV_GPU_TEMP{Hostname=~\"$Hostname\"}",
+          "expr": "DCGM_FI_DEV_GPU_TEMP{Hostname=~\"$Hostname\", namespace=~\"$namespace|\"}",
           "legendFormat": "{{Hostname}}/{{gpu}}",
           "refId": "A"
         }
@@ -751,7 +751,7 @@
       "id": 41,
       "targets": [
         {
-          "expr": "max by (Hostname, gpu) (DCGM_FI_DEV_XID_ERRORS{Hostname=~\"$Hostname\"})",
+          "expr": "max by (Hostname, gpu) (DCGM_FI_DEV_XID_ERRORS{Hostname=~\"$Hostname\", namespace=~\"$namespace|\"})",
           "legendFormat": "{{Hostname}}/{{gpu}}",
           "refId": "A"
         }
@@ -809,7 +809,7 @@
       "id": 42,
       "targets": [
         {
-          "expr": "rate(DCGM_FI_DEV_POWER_VIOLATION{Hostname=~\"$Hostname\"}[5m])",
+          "expr": "rate(DCGM_FI_DEV_POWER_VIOLATION{Hostname=~\"$Hostname\", namespace=~\"$namespace|\"}[5m])",
           "legendFormat": "{{Hostname}}/{{gpu}}",
           "refId": "A"
         }
@@ -857,7 +857,7 @@
       "id": 43,
       "targets": [
         {
-          "expr": "rate(DCGM_FI_DEV_THERMAL_VIOLATION{Hostname=~\"$Hostname\"}[5m])",
+          "expr": "rate(DCGM_FI_DEV_THERMAL_VIOLATION{Hostname=~\"$Hostname\", namespace=~\"$namespace|\"}[5m])",
           "legendFormat": "{{Hostname}}/{{gpu}}",
           "refId": "A"
         }