Nvidia DCGM Exporter インテグレーションのメトリクス
Nvidia DCGM Exporter の監視に使う Datadog の公式インテグレーションです。収集するメトリクスは 45 件あります。
公式の紹介文 (英語)
Monitors the exposed GPU metrics leveraged by the Nvidia DCGM Exporter
種別の内訳
- gauge (その時点の値): 37 件
- count (集計間隔内の発生回数): 8 件
メトリクスの一覧
説明は公式の定義の英文をそのまま載せています。
| メトリクス | 種別 | 単位 |
|---|---|---|
dcgm.Current clock throttle reasons (bitmask of DCGM_CLOCKS_THROTTLE_REASON_*) | gauge | - |
dcgm.Number of remapped rows for correctable errors. | count | row |
dcgm.Decoder utilization (in %). | gauge | percent |
dcgm.Change in number Devices on the node. | count | device |
dcgm.Number of Devices on the node. | gauge | - |
dcgm.Ratio of cycles the device memory interface is active sending or receiving data (in %). | gauge | fraction |
dcgm.Encoder utilization (in %). | gauge | percent |
dcgm.Fan speed for the device in percent 0-100. | gauge | percent |
dcgm.Free Frame Buffer in MB. | gauge | megabyte |
dcgm.Reserved Frame Buffer in MB. | gauge | megabyte |
dcgm.Total Frame Buffer of the GPU in MB. | gauge | megabyte |
dcgm.Used Frame Buffer in MB. | gauge | megabyte |
dcgm.Percentage used of Frame Buffer: Used/(Total - Reserved). Range 0.0-1.0 | gauge | fraction |
dcgm.GPU utilization (in %). | gauge | percent |
dcgm.Ratio of time the graphics engine is active (in %). | gauge | fraction |
dcgm.Memory clock frequency (in MHz). | gauge | megahertz |
dcgm.Memory utilization (in %). | gauge | percent |
dcgm.Memory temperature (in C). | gauge | degree celsius |
dcgm.Change in number of NVLink bandwidth counters for all lanes | count | - |
dcgm.Total number of NVLink bandwidth counters for all lanes | gauge | - |
dcgm.Change in number of PCIe retries. | count | - |
dcgm.Total number of PCIe retries. | gauge | - |
dcgm.Change in PCIe Rx utilization information. | count | - |
dcgm.PCIe Rx utilization information. | gauge | - |
dcgm.Change PCIe Tx utilization information. | count | - |
dcgm.PCIe Tx utilization information | gauge | - |
dcgm.Ratio of cycles the fp16 pipes are active (in %). | gauge | fraction |
dcgm.Ratio of cycles the fp32 pipes are active (in %). | gauge | fraction |
dcgm.Ratio of cycles the fp64 pipes are active (in %). | gauge | fraction |
dcgm.Ratio of cycles the tensor (HMMA) pipe is active (in %). | gauge | fraction |
dcgm.Current power limit for the device. | gauge | watt |
dcgm.Power draw (in W). | gauge | watt |
dcgm.Performance state (P-State) 0-15. 0=highest | gauge | - |
dcgm.Whether remapping of rows has failed. | gauge | - |
dcgm.Slowdown temperature for the device. | gauge | degree celsius |
dcgm.The ratio of cycles an SM has at least 1 warp assigned (in %). | gauge | fraction |
dcgm.SM clock frequency (in MHz). | gauge | megahertz |
dcgm.The ratio of number of warps resident on an SM (in %). | gauge | fraction |
dcgm.GPU temperature (in C). | gauge | degree celsius |
dcgm.Change in energy consumption (in mJ). | count | millijoule |
dcgm.Total energy consumption since boot (in mJ) | gauge | - |
dcgm.Change in number of remapped rows for uncorrectable errors. | count | row |
dcgm.Total number of remapped rows for uncorrectable errors. | gauge | - |
dcgm.vGPU License status | gauge | - |
dcgm.Value of the last XID error encountered. | gauge | - |
公式ドキュメント
メトリクスの名前と説明は Datadog の公式インテグレーションの定義 (Copyright Datadog, Inc.・BSD-3-Clause) を英文のまま載せています。利用条件の全文は サイトについてにあります。