Java QueueMetrics.getAllocatedGPUs方法代码示例

import org.apache.hadoop.yarn.server.resourcemanager.scheduler.QueueMetrics; //导入方法依赖的package包/类
public UserMetricsInfo(final ResourceManager rm, final String user) {
  ResourceScheduler rs = rm.getResourceScheduler();
  QueueMetrics metrics = rs.getRootQueueMetrics();
  QueueMetrics userMetrics = metrics.getUserMetrics(user);
  this.userMetricsAvailable = false;

  if (userMetrics != null) {
    this.userMetricsAvailable = true;

    this.appsSubmitted = userMetrics.getAppsSubmitted();
    this.appsCompleted = userMetrics.getAppsCompleted();
    this.appsPending = userMetrics.getAppsPending();
    this.appsRunning = userMetrics.getAppsRunning();
    this.appsFailed = userMetrics.getAppsFailed();
    this.appsKilled = userMetrics.getAppsKilled();

    this.runningContainers = userMetrics.getAllocatedContainers();
    this.pendingContainers = userMetrics.getPendingContainers();
    this.reservedContainers = userMetrics.getReservedContainers();

    this.reservedMB = userMetrics.getReservedMB();
    this.pendingMB = userMetrics.getPendingMB();
    this.allocatedMB = userMetrics.getAllocatedMB();

    this.reservedVirtualCores = userMetrics.getReservedVirtualCores();
    this.pendingVirtualCores = userMetrics.getPendingVirtualCores();
    this.allocatedVirtualCores = userMetrics.getAllocatedVirtualCores();

    this.reservedGPUs= userMetrics.getReservedGPUs();
    this.pendingGPUs = userMetrics.getPendingGPUs();
    this.allocatedGPUs = userMetrics.getAllocatedGPUs();
  }
}
 

import org.apache.hadoop.yarn.server.resourcemanager.scheduler.QueueMetrics; //导入方法依赖的package包/类
public ClusterMetricsInfo(final ResourceManager rm) {
  ResourceScheduler rs = rm.getResourceScheduler();
  QueueMetrics metrics = rs.getRootQueueMetrics();
  ClusterMetrics clusterMetrics = ClusterMetrics.getMetrics();

  this.appsSubmitted = metrics.getAppsSubmitted();
  this.appsCompleted = metrics.getAppsCompleted();
  this.appsPending = metrics.getAppsPending();
  this.appsRunning = metrics.getAppsRunning();
  this.appsFailed = metrics.getAppsFailed();
  this.appsKilled = metrics.getAppsKilled();

  this.reservedMB = metrics.getReservedMB();
  this.availableMB = metrics.getAvailableMB();
  this.allocatedMB = metrics.getAllocatedMB();

  this.reservedVirtualCores = metrics.getReservedVirtualCores();
  this.availableVirtualCores = metrics.getAvailableVirtualCores();
  this.allocatedVirtualCores = metrics.getAllocatedVirtualCores();

  this.reservedGPUs = metrics.getReservedGPUs();
  this.availableGPUs = metrics.getAvailableGPUs();
  this.allocatedGPUs = metrics.getAllocatedGPUs();

  this.containersAllocated = metrics.getAllocatedContainers();
  this.containersPending = metrics.getPendingContainers();
  this.containersReserved = metrics.getReservedContainers();

  if (rs instanceof CapacityScheduler) {
    this.totalMB = availableMB + allocatedMB + reservedMB;
    this.totalVirtualCores = availableVirtualCores + allocatedVirtualCores
        + containersReserved;
    this.totalGPUs = availableGPUs + allocatedGPUs + reservedGPUs;
  } else {
    this.totalMB = availableMB + allocatedMB;
    this.totalVirtualCores = availableVirtualCores + allocatedVirtualCores;
    this.totalGPUs = availableGPUs + allocatedGPUs;
  }
  this.activeNodes = clusterMetrics.getNumActiveNMs();
  this.lostNodes = clusterMetrics.getNumLostNMs();
  this.unhealthyNodes = clusterMetrics.getUnhealthyNMs();
  this.decommissioningNodes = clusterMetrics.getNumDecommissioningNMs();
  this.decommissionedNodes = clusterMetrics.getNumDecommisionedNMs();
  this.rebootedNodes = clusterMetrics.getNumRebootedNMs();
  this.shutdownNodes = clusterMetrics.getNumShutdownNMs();
  this.totalNodes = activeNodes + lostNodes + decommissionedNodes
      + rebootedNodes + unhealthyNodes + shutdownNodes;
}
 

import org.apache.hadoop.yarn.server.resourcemanager.scheduler.QueueMetrics; //导入方法依赖的package包/类
private float getPercenUsedGpus(QueueMetrics metrics) {
  int totalGpus = metrics.getAllocatedGPUs() + metrics.getAvailableGPUs();
  int usedGpus = metrics.getAllocatedGPUs() + metrics.getPendingGPUs();
  return (float) usedGpus / totalGpus;
}
 

import org.apache.hadoop.yarn.server.resourcemanager.scheduler.QueueMetrics; //导入方法依赖的package包/类
public void verifyClusterMetrics(int submittedApps, int completedApps,
    int reservedMB, int availableMB, int allocMB, int reservedVirtualCores,
    int availableVirtualCores, int allocVirtualCores, int totalVirtualCores,
    int reservedGPUs, int availableGPUs, int allocGPUs, int totalGPUs,
    int containersAlloc, int totalMB, int totalNodes, int lostNodes,
    int unhealthyNodes, int decommissionedNodes, int rebootedNodes,
    int activeNodes, int shutdownNodes) throws JSONException, Exception {

  ResourceScheduler rs = rm.getResourceScheduler();
  QueueMetrics metrics = rs.getRootQueueMetrics();
  ClusterMetrics clusterMetrics = ClusterMetrics.getMetrics();

  long totalMBExpect = 
      metrics.getAvailableMB() + metrics.getAllocatedMB();
  long totalVirtualCoresExpect = 
      metrics.getAvailableVirtualCores() + metrics.getAllocatedVirtualCores();
  long totalGPUsExpect =
      metrics.getAvailableGPUs() + metrics.getAllocatedGPUs();
  assertEquals("appsSubmitted doesn't match", 
      metrics.getAppsSubmitted(), submittedApps);
  assertEquals("appsCompleted doesn't match", 
      metrics.getAppsCompleted(), completedApps);
  assertEquals("reservedMB doesn't match",
      metrics.getReservedMB(), reservedMB);
  assertEquals("availableMB doesn't match", 
      metrics.getAvailableMB(), availableMB);
  assertEquals("allocatedMB doesn't match", 
      metrics.getAllocatedMB(), allocMB);
  assertEquals("reservedVirtualCores doesn't match",
      metrics.getReservedVirtualCores(), reservedVirtualCores);
  assertEquals("availableVirtualCores doesn't match",
      metrics.getAvailableVirtualCores(), availableVirtualCores);
  assertEquals("allocatedVirtualCores doesn't match",
      totalVirtualCoresExpect, allocVirtualCores);
  assertEquals("reservedGPUs doesn't match",
      metrics.getReservedGPUs(), reservedGPUs);
  assertEquals("availableGPUs doesn't match",
      metrics.getAvailableGPUs(), availableGPUs);
  assertEquals("allocatedGPUs doesn't match",
      totalGPUsExpect, allocGPUs);
  assertEquals("containersAllocated doesn't match", 0, containersAlloc);
  assertEquals("totalMB doesn't match", totalMBExpect, totalMB);
  assertEquals(
      "totalNodes doesn't match",
      clusterMetrics.getNumActiveNMs() + clusterMetrics.getNumLostNMs()
          + clusterMetrics.getNumDecommisionedNMs()
          + clusterMetrics.getNumRebootedNMs()
          + clusterMetrics.getUnhealthyNMs(), totalNodes);
  assertEquals("lostNodes doesn't match", clusterMetrics.getNumLostNMs(),
      lostNodes);
  assertEquals("unhealthyNodes doesn't match",
      clusterMetrics.getUnhealthyNMs(), unhealthyNodes);
  assertEquals("decommissionedNodes doesn't match",
      clusterMetrics.getNumDecommisionedNMs(), decommissionedNodes);
  assertEquals("rebootedNodes doesn't match",
      clusterMetrics.getNumRebootedNMs(), rebootedNodes);
  assertEquals("activeNodes doesn't match", clusterMetrics.getNumActiveNMs(),
      activeNodes);
  assertEquals("shutdownNodes doesn't match",
      clusterMetrics.getNumShutdownNMs(), shutdownNodes);
}