[MAPREDUCE-4843] When using DefaultTaskController, JobLocalizer not thread safe

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.1.1
• Fix Version/s: 1.2.0
• Component/s: tasktracker
• Target Version/s: 1.1.1
• Hadoop Flags: Reviewed
• Assignee: Karthik Kambatla
• Reporter: yunjiong zhao
• Created: 04/Dec/12 13:39
• Updated: 03/Nov/14 18:05
• Resolved: 04/Feb/13 22:27

Description

In our cluster, some times job will failed due to below exception:
2012-12-03 23:11:54,811 WARN org.apache.hadoop.mapred.TaskTracker: Error initializing attempt_201212031626_1115_r_000023_0:
org.apache.hadoop.util.DiskChecker$DiskErrorException: Could not find taskTracker/$username/jobcache/job_201212031626_1115/job.xml in any of the configured local directories
at org.apache.hadoop.fs.LocalDirAllocator$AllocatorPerContext.getLocalPathToRead(LocalDirAllocator.java:424)
at org.apache.hadoop.fs.LocalDirAllocator.getLocalPathToRead(LocalDirAllocator.java:160)
at org.apache.hadoop.mapred.TaskTracker.initializeJob(TaskTracker.java:1175)
at org.apache.hadoop.mapred.TaskTracker.localizeJob(TaskTracker.java:1058)
at org.apache.hadoop.mapred.TaskTracker.startNewTask(TaskTracker.java:2213)

The root cause is JobLocalizer is not thread safe.
In DefaultTaskController.initializeJob method:
JobLocalizer localizer = new JobLocalizer((JobConf)getConf(), user, jobid);
but in JobLocalizer, it just simply keep the reference of the conf.
When two TaskLauncher threads(mapLauncher and reduceLauncher) try to initializeJob at same time, it will have two JobLocalizer, but only one conf instance.
So some times ttConf.setStrings(JOB_LOCAL_CTXT, localDirs) will reset previous job's conf.
Then it will cause the previous job's job.xml stored at another user's dir.

Issue Links

• is related to: MAPREDUCE-4964 JobLocalizer#localizeJobFiles can potentially write job.xml to the wrong user's directory