container exit code 127

yarn啟動spark進行測試時候發現，幾台機器上啟動executor都失敗了，日志如下2018-11-22 14:32:24 WARN YarnAllocator:66 - Container marked as failed: container_1516236189600_0229_01_000011 on host: cloud3. Exit status: 127. Diagnostics: Exception from container-launch.

Container id: container_1516236189600_0229_01_000011
Exit code: 127
Stack trace: ExitCodeException exitCode=127: 
	at org.apache.hadoop.util.Shell.runCommand(Shell.java:601)
	at org.apache.hadoop.util.Shell.run(Shell.java:504)
	at org.apache.hadoop.util.Shell$ShellCommandExecutor.execute(Shell.java:786)
	at org.apache.hadoop.yarn.server.nodemanager.DefaultContainerExecutor.launchContainer(DefaultContainerExecutor.java:213)
	at org.apache.hadoop.yarn.server.nodemanager.containermanager.launcher.ContainerLaunch.call(ContainerLaunch.java:302)
	at org.apache.hadoop.yarn.server.nodemanager.containermanager.launcher.ContainerLaunch.call(ContainerLaunch.java:82)
	at java.util.concurrent.FutureTask.run(FutureTask.java:266)
	at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
	at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
	at java.lang.Thread.run(Thread.java:748)


Container exited with a non-zero exit code 127

2018-11-22 14:32:26 INFO  YarnAllocator:54 - Driver requested a total number of 0 executor(s).

在google上查了很久也沒解決問題，直到看到有個人建議用yarn logs -applicationId <APP_ID>查日志才發現如下情況

Container: container_1516236189600_0231_01_000001 on cloud3_22681
===================================================================
LogType:stderr
Log Upload Time:星期四 十一月 22 14:48:13 +0800 2018
LogLength:75
Log Contents:
/bin/bash: /data/platform/jdk1.8.0_144/bin/java: No such file or directory

LogType:stdout
Log Upload Time:星期四 十一月 22 14:48:13 +0800 2018
LogLength:0
Log Contents:

  已經很明顯了，然后連到相應機器上看，確實沒有jdk1.8.0_144，增加jdk1.8.0_144。重啟job發現沒問題了。