Reduce scraped and exported metrics

castai · Aug 30, 2024 · cd73184 · cd73184
1 parent 6aea75f
commit cd73184
Show file tree

Hide file tree

Showing 2 changed files with 19 additions and 27 deletions.
diff --git a/charts/gpu-metrics-exporter/templates/dcgm-exporter-configmap.yaml b/charts/gpu-metrics-exporter/templates/dcgm-exporter-configmap.yaml
@@ -6,14 +6,14 @@ metadata:
 data:
   counters.csv: |
     # Temperature and power usage,,
-    DCGM_FI_DEV_GPU_TEMP, gauge, Current temperature readings for the device in degrees C.
-    DCGM_FI_DEV_MEMORY_TEMP, gauge, Memory temperature for the device.
-    DCGM_FI_DEV_POWER_USAGE, gauge, Power usage for the device in Watts.
+    # DCGM_FI_DEV_GPU_TEMP, gauge, Current temperature readings for the device in degrees C.
+    # DCGM_FI_DEV_MEMORY_TEMP, gauge, Memory temperature for the device.
+    # DCGM_FI_DEV_POWER_USAGE, gauge, Power usage for the device in Watts.
 
     # Utilization,,
     DCGM_FI_PROF_SM_ACTIVE, gauge, The ratio of cycles an SM has at least 1 warp assigned
     DCGM_FI_PROF_SM_OCCUPANCY, gauge, The fraction of resident warps on a multiprocessor
-    DCGM_FI_PROF_PIPE_TENSOR_ACTIVE, gauge, The ratio of cycles the tensor (HMMA) pipe is active (off the peak sustained elapsed cycles)
+    # DCGM_FI_PROF_PIPE_TENSOR_ACTIVE, gauge, The ratio of cycles the tensor (HMMA) pipe is active (off the peak sustained elapsed cycles)
     DCGM_FI_PROF_GR_ENGINE_ACTIVE, gauge, Ratio of time the graphics engine is active (in %).
     DCGM_FI_PROF_DRAM_ACTIVE, gauge, The ratio of cycles the device memory interface is active sending or receiving data.
 
@@ -23,10 +23,10 @@ data:
     DCGM_FI_DEV_FB_TOTAL, gauge, Total Frame Buffer of the GPU in MB.
 
     # PCIE,,
-    DCGM_FI_PROF_PCIE_TX_BYTES, gauge, Total number of bytes transmitted through PCIe TX
-    DCGM_FI_PROF_PCIE_RX_BYTES, gauge, Total number of bytes received through PCIe RX
-    DCGM_FI_DEV_PCIE_LINK_GEN, gauge, PCIe Current Link Generation.
-    DCGM_FI_DEV_PCIE_LINK_WIDTH, gauge, PCIe Current Link Width.
+    # DCGM_FI_PROF_PCIE_TX_BYTES, gauge, Total number of bytes transmitted through PCIe TX
+    # DCGM_FI_PROF_PCIE_RX_BYTES, gauge, Total number of bytes received through PCIe RX
+    # DCGM_FI_DEV_PCIE_LINK_GEN, gauge, PCIe Current Link Generation.
+    # DCGM_FI_DEV_PCIE_LINK_WIDTH, gauge, PCIe Current Link Width.
 
     # Pipelines,,
     DCGM_FI_PROF_PIPE_INT_ACTIVE, gauge, Ratio of cycles the integer pipe is active.

diff --git a/internal/exporter/types.go b/internal/exporter/types.go
@@ -26,24 +26,16 @@ const (
 
 var (
 	EnabledMetrics = map[MetricName]struct{}{
-		MetricStreamingMultiProcessorActive:       {},
-		MetricStreamingMultiProcessorOccupancy:    {},
-		MetricStreamingMultiProcessorTensorActive: {},
-		MetricDRAMActive:                          {},
-		MetricPCIeTXBytes:                         {},
-		MetricPCIeRXBytes:                         {},
-		MetricGraphicsEngineActive:                {},
-		MetricFrameBufferTotal:                    {},
-		MetricFrameBufferFree:                     {},
-		MetricFrameBufferUsed:                     {},
-		MetricPCIeLinkGen:                         {},
-		MetricPCIeLinkWidth:                       {},
-		MetricGPUTemperature:                      {},
-		MetricMemoryTemperature:                   {},
-		MetricPowerUsage:                          {},
-		MetricIntPipeActive:                       {},
-		MetricFloat16PipeActive:                   {},
-		MetricFloat32PipeActive:                   {},
-		MetricFloat64PipeActive:                   {},
+		MetricGraphicsEngineActive:             {},
+		MetricFrameBufferTotal:                 {},
+		MetricFrameBufferFree:                  {},
+		MetricFrameBufferUsed:                  {},
+		MetricStreamingMultiProcessorActive:    {},
+		MetricStreamingMultiProcessorOccupancy: {},
+		MetricDRAMActive:                       {},
+		MetricIntPipeActive:                    {},
+		MetricFloat16PipeActive:                {},
+		MetricFloat32PipeActive:                {},
+		MetricFloat64PipeActive:                {},
 	}
 )