File size: 4,614 Bytes
fb0bedf |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 |
cluster_metric_header = [
"NODE_CPU_UTILIZATION",
"NODE_MEMORY_UTILIZATION",
"NODE_IB_SEND",
"NODE_IB_RECEIVE",
]
dcgm_metric_header = [
"XID_ERRORS",
"GPU_TEMP",
"MEMORY_TEMP",
"MEM_CLOCK",
"MEM_COPY_UTIL",
"FB_FREE",
"FB_USED",
"DRAM_ACTIVE",
"POWER_USAGE",
"GPU_UTIL",
"PIPE_TENSOR_ACTIVE",
"SM_ACTIVE",
"SM_OCCUPANCY",
]
dcgm_full = [
"DCGM_FI_DEV_CORRECTABLE_REMAPPED_ROWS", # 可纠正错误的重新映射行数 Number of remapped rows for correctable errors.
"DCGM_FI_DEV_DEC_UTIL", # GPU 解码器利用率（单位：%）
"DCGM_FI_DEV_ECC_DBE_AGG_TOTAL", # 双位聚合（持久）ECC 错误总数 (Total double bit aggregate (persistent) ECC errors) 注：单调递增。
"DCGM_FI_DEV_ECC_DBE_VOL_TOTAL", # 双位易失性 ECC 错误总数。(Total double bit volatile ECC errors)
"DCGM_FI_DEV_ECC_SBE_AGG_TOTAL", # 单位聚合（持久）ECC 错误总数 注：单调递增。
"DCGM_FI_DEV_ECC_SBE_VOL_TOTAL", # 单位易失性 ECC 错误总数。
"DCGM_FI_DEV_ENC_UTIL", # GPU 编码器利用率（单位：%）
"DCGM_FI_DEV_FB_FREE", # GPU 帧缓存剩余量（单位：MiB）
"DCGM_FI_DEV_FB_USED", # GPU 帧缓存使用量（单位：MiB） (该值与nvidia-smi命令中memory-usage的已使用值对应)
"DCGM_FI_DEV_GPU_TEMP", # GPU 当前温度（单位：℃）
"DCGM_FI_DEV_GPU_UTIL", # GPU 利用率（单位：%）
"DCGM_FI_DEV_MEMORY_TEMP", # 显存当前温度（单位：℃）
"DCGM_FI_DEV_MEM_CLOCK", # GPU 内存时钟（单位：MHZ）
"DCGM_FI_DEV_MEM_COPY_UTIL", # GPU 内存带宽利用率（单位：%）
"DCGM_FI_DEV_NVLINK_BANDWIDTH_TOTAL", # GPU 所有通道的 NVLink 带宽计数器总数
"DCGM_FI_DEV_PCIE_REPLAY_COUNTER", # GPU PCIE 重试次数
"DCGM_FI_DEV_POWER_USAGE", # GPU 当前使用功率（单位：W)
"DCGM_FI_DEV_ROW_REMAP_FAILURE", # Whether remapping of rows has failed
"DCGM_FI_DEV_SM_CLOCK", # GPU SM 时钟（单位：MHZ）
"DCGM_FI_DEV_TOTAL_ENERGY_CONSUMPTION", # GPU 启动以来的总能耗（单位：mJ）
"DCGM_FI_DEV_UNCORRECTABLE_REMAPPED_ROWS", # Number of remapped rows for uncorrectable errors.
"DCGM_FI_DEV_VGPU_LICENSE_STATUS", # vGPU 许可证状态
"DCGM_FI_DEV_XID_ERRORS", # XID errors
"DCGM_FI_PROF_DRAM_ACTIVE", # 内存接口主动发送或接收数据的周期比率 (The ratio of cycles the device memory interface is active sending or receiving data)
"DCGM_FI_PROF_GR_ENGINE_ACTIVE", # Profiling Fields. These all start with DCGM_FI_PROF_* Ratio of time the graphics engine is active. The graphics engine is active if a graphics/compute context is bound and the graphics pipe or compute pipe is busy.
"DCGM_FI_PROF_PCIE_RX_BYTES", # GPU PCIE 接收字节总数（单位：字节）
"DCGM_FI_PROF_PCIE_TX_BYTES", # GPU PCIE 发送字节总数（单位：字节）
"DCGM_FI_PROF_PIPE_TENSOR_ACTIVE", # The ratio of cycles the any tensor pipe is active (off the peak sustained elapsed cycles)
"DCGM_FI_PROF_SM_ACTIVE", # The ratio of cycles an SM has at least 1 warp assigned (computed from the number of cycles and elapsed cycles)
"DCGM_FI_PROF_SM_OCCUPANCY", # The ratio of number of warps resident on an SM.
]
table_dcgm_fatal_header = [
"DCGM_FI_DEV_XID_ERRORS",
"DCGM_FI_DEV_GPU_TEMP",
"DCGM_FI_DEV_MEMORY_TEMP",
"DCGM_FI_DEV_MEM_CLOCK",
"DCGM_FI_DEV_MEM_COPY_UTIL",
"DCGM_FI_DEV_FB_FREE",
"DCGM_FI_DEV_FB_USED",
"DCGM_FI_PROF_DRAM_ACTIVE",
"DCGM_FI_DEV_POWER_USAGE",
"DCGM_FI_DEV_GPU_UTIL",
"DCGM_FI_PROF_PIPE_TENSOR_ACTIVE",
"DCGM_FI_PROF_SM_ACTIVE",
"DCGM_FI_PROF_SM_OCCUPANCY",
"DCGM_FI_DEV_SM_CLOCK",
]
converted_dcgm_fatal_table_header = [
"XID_ERRORS",
"GPU_TEMP",
"MEMORY_TEMP",
"MEM_CLOCK",
"MEM_COPY_UTIL",
"FB_FREE",
"FB_USED",
"DRAM_ACTIVE",
"POWER_USAGE",
"GPU_UTIL",
"PIPE_TENSOR_ACTIVE",
"SM_ACTIVE",
"SM_OCCUPANCY",
"SM_CLOCK",
]
# node_infiniband_physical_state_id:
# (0: no change, 1: sleep, 2: polling, 3: disable, 4: shift, 5: link up, 6: link error recover, 7: phytest)
IB_PHYSICAL_STATE_MAP = {
"0": "no change",
"1": "sleep",
"2": "polling",
"3": "disable",
"4": "shift",
"5": "link up",
"6": "link error recover",
"7": "phytest",
}
# node_infiniband_state_id: (0: no change, 1: down, 2: init, 3: armed, 4: active, 5: act defer)
IB_STATE_MAP = {"0": "no change", "1": "down", "2": "init", "3": "armed", "4": "active", "5": "act defer"}
|