[
https://issues.apache.org/jira/browse/HADOOP-4854?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12656882#action_12656882
]
Vivek Ratan commented on HADOOP-4854:
-------------------------------------
Looks good. Some minor comments.
# In _TestCapacityScheduler.java_:
#* _FakeResourceManagerConf.reclaimCapcityInterval_, the new member variable
you've introduced, is misspelt (missing an 'a' in 'Capacity'). It should be
_reclaimCapacityInterval_.
#* As I commented earlier, remove the code in setup(int, int, int) which
explicitly sets "mapred.capacity-scheduler.reclaimCapacity.interval" in the
configuration. This call is unnecessary. Remove the followling lines:
{code}
// set interval to a large number so thread doesn't interfere with us
conf.setLong("mapred.capacity-scheduler.reclaimCapacity.interval", 500);
{code}
# In _xdocs/capacity_scheduler.xml_, use the following writeup:
{noformat}
<section>
<title>Configuring the capacity scheduler</title>
<p>The capacity scheduler's behavior can be controlled through the
following properties.
</p>
<table>
<tr>
<th>Name</th><th>Description</th>
</tr>
<tr>
<td>mapred.capacity-scheduler.reclaimCapacity.interval</td>
<td>The time interval, in seconds, between which the scheduler
periodically determines whether capacity needs to be reclaimed
for any queue. The default value is 5 seconds.
</td>
</tr>
</table>
</section>
{noformat}
# In _conf/capacity-scheduler.xml.template_, use the following writeup:
{noformat}
<property>
<name>mapred.capacity-scheduler.reclaimCapacity.interval</name>
<value>5</value>
<description>The time interval, in seconds, between which the scheduler
periodically determines whether capacity needs to be reclaimed for any
queue.
</description>
</property>
{noformat}
> Capacity Scheduler should read the reclaim-capacity-interval config value
> from its own config file
> --------------------------------------------------------------------------------------------------
>
> Key: HADOOP-4854
> URL: https://issues.apache.org/jira/browse/HADOOP-4854
> Project: Hadoop Core
> Issue Type: Bug
> Components: contrib/capacity-sched
> Reporter: Vivek Ratan
> Assignee: Sreekanth Ramakrishnan
> Priority: Minor
> Attachments: HADOOP-4854-1.patch, HADOOP-4854-2.patch,
> HADOOP-4854-3.patch
>
>
> The Capacity Scheduler needs to know the time interval for checking whether
> capacity needs to be reclaimed. This is represented by the static variable
> RECLAIM_CAPACITY_INTERVAL. This value is currently read as follows:
> {code}
> RECLAIM_CAPACITY_INTERVAL =
> conf.getLong("mapred.capacity-scheduler.reclaimCapacity.interval", 5);
> {code}
> This is incorrect for the following reasons:
> * ""mapred.capacity-scheduler.reclaimCapacity.interval" is not present in any
> of the config files: hadoop-site.xml or capacity-scheduler.xml.
> * "mapred.capacity-scheduler.reclaimCapacity.interval" should be specified in
> the scheduler's config file, capacity-scheduler.xml. It should be read
> through org.apache.hadoop.mapred.CapacitySchedulerConf.
> * The Forrest documentation for the Capacity Scheduler should describe this
> setting. We should probably create a new sub-section under 'Configuration',
> titled 'Configuring properties for the Scheduler'.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
