[
https://issues.apache.org/jira/browse/HADOOP-4472?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12641761#action_12641761
]
Hemanth Yamijala commented on HADOOP-4472:
------------------------------------------
If we choose to make the schedulers unaware of the setup/cleanup tasks, one
more consideration is whether the JT should add the job to the scheduler only
after setup task is complete.
Consider this example: two jobs are added J1 and J2. Say J1's setup task takes
a long time to complete, and J2's setup task completes very soon. If the
scheduler is informed about the two jobs before their setup tasks have
completed (as it happens today), the scheduler would look at J1 and initialize
its tasks. But since setup has not completed, it would move to initialize the
second job as well. Initializing the first job at this point seems to be
wastefully occupying JT memory, as no M/R task can run until setup is complete.
Put another way, letting task initialization happen only after setup has
completed will help to reduce the memory footprint of the JT.
The flip side of this approach is that tasks would be given to the scheduler in
order of completion of the setup tasks (irrespective of other aspects like
submission time, priority, etc). However, I believe at some point (in 0.19)
setup was being done as part of the job client, before they were graduated to
tasks. That implicitly meant that jobs are given to the scheduler in order of
completion of setup tasks. So, it may not be that bad.
Does this make sense ?
> Should we move out the creation of setup/cleanup tasks from
> JobInProgress.initTasks()?
> ---------------------------------------------------------------------------------------
>
> Key: HADOOP-4472
> URL: https://issues.apache.org/jira/browse/HADOOP-4472
> Project: Hadoop Core
> Issue Type: Improvement
> Components: mapred
> Reporter: Vivek Ratan
>
> JobInProgress.initTasks() creates TIPs for map and reduce tasks, and also the
> newly-introduced setup and cleanup tasks. initTasks() is called by the
> schedulers, as for reasons of memory optimizations, schedulers may choose to
> initialize M/R tasks at various moments (the Capacity Scheduler, for example,
> calls initTasks() just when it considers a job for running). One can say that
> Schedulers 'own' the initialization of M/R tasks in a job. Furthermore the JT
> 'owns' the setup and cleanup tasks (it schedules them, and Schedulers are
> unaware of these tasks). This causes a problematic dependency between the JT
> and a Scheduler. For example, the Capacity Scheduler calls initTasks() and
> immediately calls JobInProgress.obtainNewMapTask for a map task. This is a
> problem today, because we cannot run any map or reduce tasks before the setup
> task is run, which the Capacity Scheduler is not aware of.
> Either all Schedulers are explicitly aware of setup/cleanup tasks and their
> dependencies with M/R tasks (in which case, Schedulers 'own' the creation and
> scheduling of all these tasks correctly), or the JT 'owns' the setup/cleanup
> tasks and Schedulers are completely unaware of them (in which case, the
> creation of setup/cleanup tasks must be moved out of initTasks into a
> separate method which is called by the JT).
> I think the latter is the right way to go (unless we implement HADOOP-4421,
> in which case the former option may be viable as well).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
