Device_utilization_desc_of_container 数据采集始终为0

[MouseSun846]

Device_utilization_desc_of_container 数据采集始终为0， 调试libvgpu 发现找不到hostpid

代码修改如下

int get_used_gpu_utilization(int *userutil,int *sysprocnum) {
    struct timeval cur;
    size_t microsec;

    int i,sum=0;
    unsigned int infcount;
    size_t summonitor=0;
    nvmlProcessInfo_v1_t infos[SHARED_REGION_MAX_PROCESS_NUM];

    unsigned int nvmlCounts;
    CHECK_NVML_API(nvmlDeviceGetCount(&nvmlCounts));

    int devi,cudadev;
    for (devi=0;devi<nvmlCounts;devi++){
      sum=0;
      summonitor=0;
      shrreg_proc_slot_t *proc;
      cudadev = nvml_to_cuda_map((unsigned int)(devi));
      if (cudadev<0)
        continue;
      nvmlDevice_t device;
      char uuid[NVML_DEVICE_UUID_BUFFER_SIZE];
      CHECK_NVML_API(nvmlDeviceGetHandleByIndex(cudadev, &device));
        // Get device UUID
     CHECK_NVML_API(nvmlDeviceGetUUID(device, uuid, NVML_DEVICE_UUID_BUFFER_SIZE));
      nvmlReturn_t res = nvmlDeviceGetComputeRunningProcesses(device,&infcount,infos);
      if (res==NVML_ERROR_INSUFFICIENT_SIZE){
        continue;
      }
      gettimeofday(&cur,NULL);
      microsec = (cur.tv_sec - 1) * 1000UL * 1000UL + cur.tv_usec;
      nvmlProcessUtilizationSample_t processes_sample[SHARED_REGION_MAX_PROCESS_NUM];
      unsigned int processes_num = SHARED_REGION_MAX_PROCESS_NUM;
      res = nvmlDeviceGetProcessUtilization(device,processes_sample,&processes_num,microsec);
      LOG_DEBUG("processes_num=%d\n",processes_num);
      LOG_DEBUG("Device UUID: %s\n", uuid);
      LOG_DEBUG("nvmlDeviceGetProcessUtilization res: %d\n", res);
      if ((res==NVML_ERROR_INSUFFICIENT_SIZE) || (res==NVML_ERROR_NOT_FOUND)){
        userutil[cudadev] = 0;
        for (i=0; i<infcount; i++){
          proc = find_proc_by_hostpid(infos[i].pid);
          if (proc != NULL){
              LOG_DEBUG("pid=%u monitor=%lld\n",infos[i].pid,infos[i].usedGpuMemory);
              summonitor += infos[i].usedGpuMemory;
          }
          set_gpu_device_memory_monitor(infos[i].pid,cudadev,summonitor);
          set_gpu_device_sm_utilization(infos[i].pid,cudadev,0);
        } 
        continue;
      }
      for (i=0; i<processes_num; i++){
          //if (processes_sample[i].timeStamp >= microsec){
          proc = find_proc_by_hostpid(processes_sample[i].pid);
          if (proc != NULL){
              //LOG_WARN("pid=%u num=%d\n",processes_sample[i].pid,processes_num);
              //proc = find_proc_by_hostpid(processes_sample[i].pid);
              //if (proc!=NULL) {
              //    printf("inner pid=%u\n",proc->pid);
              sum += processes_sample[i].smUtil;
              summonitor += infos[i].usedGpuMemory;
              LOG_DEBUG("monitorused=%lld %d %d %d",infos[i].usedGpuMemory,proc->hostpid,proc->pid,pidfound);
              LOG_DEBUG("smutil=%u %u %u\n",processes_sample[i].smUtil,processes_sample[i].encUtil,processes_sample[i].decUtil);
              //}
          } else {
            LOG_DEBUG("find_proc_by_hostpid proc is null.\n")
          }

          set_gpu_device_memory_monitor(processes_sample[i].pid,cudadev,summonitor);
          set_gpu_device_sm_utilization(processes_sample[i].pid,cudadev,processes_sample[i].smUtil);
      }
      if (sum < 0)
        sum = 0;
      LOG_DEBUG("sum=%d\n", sum);
      userutil[cudadev] = sum;
    }
    return 0;
}

请问是配置不正确吗？