YARN常见异常

异常1：

2012-05-16 16:18:42,468 WARN org.apache.hadoop.yarn.server.nodemanager.containermanager.launcher.ContainerLaunch: Failed to launch container.

java.io.FileNotFoundException: File /tmp/nm-local-dir/usercache/a/appcache/application_1337150856633_0016 does not exist

        at org.apache.hadoop.fs.RawLocalFileSystem.getFileStatus(RawLocalFileSystem.java:449)

        at org.apache.hadoop.fs.FileSystem.primitiveMkdir(FileSystem.java:864)

        at org.apache.hadoop.fs.DelegateToFileSystem.mkdir(DelegateToFileSystem.java:143)

        at org.apache.hadoop.fs.FilterFs.mkdir(FilterFs.java:189)

        at org.apache.hadoop.fs.FileContext$4.next(FileContext.java:700)

        at org.apache.hadoop.fs.FileContext$4.next(FileContext.java:697)

        at org.apache.hadoop.fs.FileContext$FSLinkResolver.resolve(FileContext.java:2319)

        at org.apache.hadoop.fs.FileContext.mkdir(FileContext.java:697)

原因A:

yarn-site.xml中参数yarn.nodemanager.local-dirs和yarn.nodemanager.log-dirs使用

默认路径（/tmp/..），导致将磁盘空间撑爆

原因B:

在org.apache.hadoop.yarn.server.nodemanager.DefaultContainerExecutor文件中，创建目录方法在默认情况下需要创建的目录父目录不存在则失败，具体代码如下

 

  @Override
  public int launchContainer(Container container,
      Path nmPrivateContainerScriptPath, Path nmPrivateTokensPath,
      String userName, String appId, Path containerWorkDir,
      List<String> localDirs, List<String> logDirs) throws IOException {

    ContainerId containerId = container.getContainerID();

    // create container dirs on all disks
    String containerIdStr = ConverterUtils.toString(containerId);
    String appIdStr =
        ConverterUtils.toString(
            container.getContainerID().getApplicationAttemptId().
                getApplicationId());
    for (String sLocalDir : localDirs) {
      Path usersdir = new Path(sLocalDir, ContainerLocalizer.USERCACHE);
      Path userdir = new Path(usersdir, userName);
      Path appCacheDir = new Path(userdir, ContainerLocalizer.APPCACHE);
      Path appDir = new Path(appCacheDir, appIdStr);
      Path containerDir = new Path(appDir, containerIdStr);
      lfs.mkdir(containerDir, null, false);
    }

 该方法第三个参数为false的时候，如果创建目录的父目录不存在，则调用失败，将其改为

 

lfs.mkdir(containerDir, null, true);

异常2：

2012-05-16 09:56:57,078 INFO org.apache.hadoop.yarn.server.nodemanager.containermanager.localizer.ResourceLocalizationService: Localizer failed

java.lang.ArithmeticException: / by zero

        at org.apache.hadoop.fs.LocalDirAllocator$AllocatorPerContext.getLocalPathForWrite(LocalDirAllocator.java:355)

        at org.apache.hadoop.fs.LocalDirAllocator.getLocalPathForWrite(LocalDirAllocator.java:150)

        at org.apache.hadoop.fs.LocalDirAllocator.getLocalPathForWrite(LocalDirAllocator.java:131)

        at org.apache.hadoop.fs.LocalDirAllocator.getLocalPathForWrite(LocalDirAllocator.java:115)

        at org.apache.hadoop.yarn.server.nodemanager.LocalDirsHandlerService.getLocalPathForWrite(LocalDirsHandlerService.java:257)

        at org.apache.hadoop.yarn.server.nodemanager.containermanager.localizer.ResourceLocalizationService$LocalizerRunner.run(ResourceLocalizationService.java:854)

 

原因：

该错误是由于无法创建本地文件而产生的，可能会由上述磁盘空间不足导致,重定向上面提到的两个参数即可

异常3：

查看log没有明显的ERROR，但存在类似以下描述的日志

2012-05-16 13:08:20,876 INFO org.apache.hadoop.yarn.server.nodemanager.NodeStatusUpdaterImpl: Sending out status for container: container_id {, app_attempt_id {, application_id {, id: 18, cluster_timestamp: 1337134318909, }, attemptId: 1, }, id: 6, }, state: C_COMPLETE, diagnostics: "Container [pid=15641,containerID=container_1337134318909_0018_01_000006] is running beyond virtual memory limits. Current usage: 32.1mb of 1.0gb physical memory used; 6.2gb of 2.1gb virtual memory used. Killing container.\nDump of the process-tree for container_1337134318909_0018_01_000006 :\n\t|- PID PPID PGRPID SESSID CMD_NAME USER_MODE_TIME(MILLIS) SYSTEM_TIME(MILLIS) VMEM_USAGE(BYTES) RSSMEM_USAGE(PAGES) FULL_CMD_LINE\n\t|- 15641 26354 15641 15641 (java) 36 2 6686339072 8207 /home/zhouchen.zm/jdk1.6.0_23/bin/java

 

原因：

该错误是YARN的虚拟内存计算方式导致，上例中用户程序申请的内存为1Gb，YARN根据此值乘以一个比例（默认为2.1）得出申请的虚拟内存的值，当YARN计算的用户程序所需虚拟内存值大于计算出来的值时，就会报出以上错误。调节比例值可以解决该问题。具体参数为：yarn-site.xml中的yarn.nodemanager.vmem-pmem-ratio