[
https://issues.apache.org/jira/browse/HADOOP-3740?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12612759#action_12612759
]
Amar Kamat commented on HADOOP-3740:
------------------------------------
I had an offline discussion with Alejandro. So the problem seems to be of
_resource constraints_. Here the resource is some service that requires a
'license'. This falls into a broad category of resource aware schedulers. So
ideally one should make use of some scheduler which is resource aware. As that
might take time we can now have a framework support for such cases where the
resources that the tasks depend upon is centrally/globally controlled. So I
think for now we can do what Alejandro suggested.
Alejandro,
Today, we dont discriminate between tasks but now we will have to do it. A
simpler approach would be to do what the scheduler does but before giving the
task in {{JobInProgress.findTaskFromList()}} and
{{JobInProgress.findSpeculativeTask()}} just check if the resource is available
or not. If yes then _return_ else _ignore and continue_. Thus we will be able
to find out an _appropriate_ task in one scheduling cycle. Updating the license
server can be done in {{JobInProgress.updateTaskStatus()}} upon task-attempt
completion (success/failure/kill).
Thoughts?
> Make JobInProgress pluggable
> ----------------------------
>
> Key: HADOOP-3740
> URL: https://issues.apache.org/jira/browse/HADOOP-3740
> Project: Hadoop Core
> Issue Type: Improvement
> Components: mapred
> Environment: all
> Reporter: Alejandro Abdelnur
>
> By allowing a pluggable JobInProgess it will be possible for provide
> implementations that can do a sophisticated task provisioning to the
> JobTracker.
> For example, by providing alternate implementations of the
> {{obtainNewMapTask}}, {{obtainNewReduceTask}} and {{updateTaskInProgress}} it
> would be possible to implement a license server that allows to throttle use
> of external resources (ie webservices, databases) so at any given time there
> are not more than N tasks using a given resource. For this a task could be
> tagged with the names of external resources and the license server would keep
> track of the tasks running per tag, if the counter reaches zero then the
> {{obtainNew*Task}} method could return NULL instead of a task.
>
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
