[
https://issues.apache.org/jira/browse/HADOOP-1819?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Owen O'Malley updated HADOOP-1819:
----------------------------------
Attachment: jt-init.patch
This patch snowballed a bit on me. It does:
1. It binds the port before clearing the system directory. This is the
primary change since it makes sure that it is being run on the right host
before wiping the system directory.
2. The error handling for wrong and unknown host are improved.
3. The HTTP server comes up before any long blocking operations happen.
4. Introduces the JobTracker state and provides it as part of ClusterStatus.
The JobTracker is INITIALIZING while cleaning up the system directory and local
storage. When it is up, the state is RUNNING. The state is visible in the
web/ui.
5. Jobs are rejected before the JobTracker is RUNNING.
6. Removes static variable that stores the JobTracker.
7. Only dump the system properties from the JobTracker at debug logging level.
8. Change the JobSubmissionProtocol to version 4.
9. Change the jsps to get the JobTracker from an attribute rather than the
static variable.
> The JobTracker should ensure that it is running on the right host.
> ------------------------------------------------------------------
>
> Key: HADOOP-1819
> URL: https://issues.apache.org/jira/browse/HADOOP-1819
> Project: Hadoop
> Issue Type: Bug
> Components: mapred
> Affects Versions: 0.14.0
> Reporter: Owen O'Malley
> Assignee: Owen O'Malley
> Fix For: 0.15.0
>
> Attachments: jt-init.patch
>
>
> We had someone start up a private JobTracker with the standard config.
> Unfortunately, it used the system directory from the config and wiped it
> clean. I want to add a sanity check that makes sure the JobTracker is being
> run on the right node before the system directory is wiped.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
