You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-dev@hadoop.apache.org by "Amareshwari Sriramadasu (JIRA)" <ji...@apache.org> on 2008/07/09 07:32:31 UTC
[jira] Commented: (HADOOP-3713) broken symlinks in jobcache when
local tasks are done but job is in progress
[ https://issues.apache.org/jira/browse/HADOOP-3713?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12611900#action_12611900 ]
Amareshwari Sriramadasu commented on HADOOP-3713:
-------------------------------------------------
On branch 0.17, when a task gets a KILLTASKACTION and the tasktracker doesnt have anymore tasks running for that job, the job is removed from runningJobs data structure. So, when it gets a task for that job again, it does localize again, which is causing the Mkdirs to fail, because the job directory is not cleanedup properly. HADOOP-3370 has fixed the problem on 0.18.
I think HADOOP-3370 should be also be applied for 0.17. That will solve this issue.
Still, there could be garbage left on the jobcache, which is addressed by HADOOP-3386.
> broken symlinks in jobcache when local tasks are done but job is in progress
> ----------------------------------------------------------------------------
>
> Key: HADOOP-3713
> URL: https://issues.apache.org/jira/browse/HADOOP-3713
> Project: Hadoop Core
> Issue Type: Bug
> Components: mapred
> Affects Versions: 0.17.0
> Reporter: Rajiv Chittajallu
> Assignee: Amareshwari Sriramadasu
> Priority: Blocker
> Fix For: 0.17.2
>
>
> When all running tasks on a tasktracker are done, not all links for /<mapred.local.dir>/taskTracker/jobcache/<job>/work are deleted. This is resulting in new tasks from the same job scheduled on this node to fail with
> 2008-07-07 17:44:49,756 INFO org.apache.hadoop.mapred.TaskTracker: LaunchTaskAction: task_200807071715_0022_r_000295_0
> 2008-07-07 17:44:49,773 WARN org.apache.hadoop.mapred.TaskTracker: Error initializing task_200807071715_0022_r_000295_0:
> java.io.IOException: Mkdirs failed to create /tmp3/taskTracker/jobcache/job_200807071715_0022/work
> at org.apache.hadoop.mapred.TaskTracker.localizeJob(TaskTracker.java:680)
> at org.apache.hadoop.mapred.TaskTracker.startNewTask(TaskTracker.java:1274)
> at org.apache.hadoop.mapred.TaskTracker.offerService(TaskTracker.java:915)
> at org.apache.hadoop.mapred.TaskTracker.run(TaskTracker.java:1310)
> at org.apache.hadoop.mapred.TaskTracker.main(TaskTracker.java:2251)
> $ ls -lt /tmp3/taskTracker/jobcache/job_200807071715_0022/work
> lrwxrwxrwx 1 user users 135 Jul 7 17:44 /tmp3/taskTracker/jobcache/job_200807071715_0022/work -> /tmp0/taskTracker/jobcache/job_200807071715_0022/work
> $ ls -lt /tmp0/mapred-local/taskTracker/jobcache/job_200807071715_0022/work
> ls: /tmp0/taskTracker/jobcache/job_200807071715_0022/work: No such file or directory
> Earlier tasks scheduled on this tasktracker have completed successfully
> 2008-07-07 17:44:44,926 INFO org.apache.hadoop.mapred.TaskRunner: task_200807071715_0022_r_000004_0 done; removing files.
> 2008-07-07 17:44:44,931 INFO org.apache.hadoop.mapred.TaskRunner: task_200807071715_0022_r_000176_0 done; removing files.
> 2008-07-07 17:44:44,958 INFO org.apache.hadoop.mapred.TaskRunner: task_200807071715_0022_r_000210_0 done; removing files.
> 2008-07-07 17:44:49,486 INFO org.apache.hadoop.mapred.TaskRunner: task_200807071715_0022_r_000153_0 done; removing files.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.