[
https://issues.apache.org/jira/browse/MAPREDUCE-1221?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12838107#action_12838107
]
Hong Tang commented on MAPREDUCE-1221:
--------------------------------------
Chip in my 2 cents. I think the goal of this jira is to prevent a node from
swapping (same as in MAPREDUCE-257). In this regard, containing memory usage
for individual tasks may not be sufficient to protect the system from swapping
- for instance, there could be foreign processes or bugs in TT and DN that "eat
up" ram on the same node and lead to swapping. Or there could be faulty ram
modules such that OS does not recognize them during the boot time and leads to
the actual amount of RAM less than configured. In my view, this idea is similar
to the OOM killer mechanism in Linux kernel and serves as a low-level
protection against faults not easily preventable from upper layers.
In terms of what tasks to shot down, when a node goes into swap, there is a
high chance that all tasks would fail (time out). Even worse, when TT is
swapping, it may not even respond to kill commands from JT. So killing some
tasks and letting others proceed may still be better than not killing. That
said, I agree that we can fine tune the policy - e.g. avoid killing the 4th
task attempt, or bias tasks that have already done lots of work (input bytes
consumed, slot secs used).
> Kill tasks on a node if the free physical memory on that machine falls below
> a configured threshold
> ---------------------------------------------------------------------------------------------------
>
> Key: MAPREDUCE-1221
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-1221
> Project: Hadoop Map/Reduce
> Issue Type: Improvement
> Components: tasktracker
> Affects Versions: 0.22.0
> Reporter: dhruba borthakur
> Assignee: Scott Chen
> Fix For: 0.22.0
>
> Attachments: MAPREDUCE-1221-v1.patch, MAPREDUCE-1221-v2.patch,
> MAPREDUCE-1221-v3.patch
>
>
> The TaskTracker currently supports killing tasks if the virtual memory of a
> task exceeds a set of configured thresholds. I would like to extend this
> feature to enable killing tasks if the physical memory used by that task
> exceeds a certain threshold.
> On a certain operating system (guess?), if user space processes start using
> lots of memory, the machine hangs and dies quickly. This means that we would
> like to prevent map-reduce jobs from triggering this condition. From my
> understanding, the killing-based-on-virtual-memory-limits (HADOOP-5883) were
> designed to address this problem. This works well when most map-reduce jobs
> are Java jobs and have well-defined -Xmx parameters that specify the max
> virtual memory for each task. On the other hand, if each task forks off
> mappers/reducers written in other languages (python/php, etc), the total
> virtual memory usage of the process-subtree varies greatly. In these cases,
> it is better to use kill-tasks-using-physical-memory-limits.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
