[
https://issues.apache.org/jira/browse/HADOOP-4665?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12674930#action_12674930
]
Hemanth Yamijala commented on HADOOP-4665:
------------------------------------------
Amr, that's a good point about the benefit to the community.
One middle ground option could be to make it part of Hadoop 0.20.1. so that the
community doesn't wait until 0.21 (which could be way out right now).
One specific concern I had about this patch (from a cursory glance) was that it
was locking the JobTracker, (the taskTrackerManager instance) in the update
method which runs frequently. In general, this could interfere with regular
processing in the jobtracker, like heartbeats (which are also synchronized on
the same instance). We've in the past seen issues of this nature on large
clusters. When the JT is locked up, tasktrackers could get lost and tasks could
fail arbitrarily. HADOOP-5280 seems to be one specific instance of this (though
we've not ascertained the reason JT got locked up).
Does the middle ground option make sense ? It may help all of us with only a
little compromise, no ?
> Add preemption to the fair scheduler
> ------------------------------------
>
> Key: HADOOP-4665
> URL: https://issues.apache.org/jira/browse/HADOOP-4665
> Project: Hadoop Core
> Issue Type: New Feature
> Components: contrib/fair-share
> Reporter: Matei Zaharia
> Fix For: 0.20.0
>
> Attachments: fs-preemption-v0.patch, hadoop-4665-v1.patch,
> hadoop-4665-v1b.patch
>
>
> Task preemption is necessary in a multi-user Hadoop cluster for two reasons:
> users might submit long-running tasks by mistake (e.g. an infinite loop in a
> map program), or tasks may be long due to having to process large amounts of
> data. The Fair Scheduler (HADOOP-3746) has a concept of guaranteed capacity
> for certain queues, as well as a goal of providing good performance for
> interactive jobs on average through fair sharing. Therefore, it will support
> preempting under two conditions:
> 1) A job isn't getting its _guaranteed_ share of the cluster for at least T1
> seconds.
> 2) A job is getting significantly less than its _fair_ share for T2 seconds
> (e.g. less than half its share).
> T1 will be chosen smaller than T2 (and will be configurable per queue) to
> meet guarantees quickly. T2 is meant as a last resort in case non-critical
> jobs in queues with no guaranteed capacity are being starved.
> When deciding which tasks to kill to make room for the job, we will use the
> following heuristics:
> - Look for tasks to kill only in jobs that have more than their fair share,
> ordering these by deficit (most overscheduled jobs first).
> - For maps: kill tasks that have run for the least amount of time (limiting
> wasted time).
> - For reduces: similar to maps, but give extra preference for reduces in the
> copy phase where there is not much map output per task (at Facebook, we have
> observed this to be the main time we need preemption - when a job has a long
> map phase and its reducers are mostly sitting idle and filling up slots).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
