[
https://issues.apache.org/jira/browse/YARN-642?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13655089#comment-13655089
]
Sandy Ryza commented on YARN-642:
---------------------------------
bq. A general suggestion: Please add a little more information to the title
than it is.
Understood
bq. Even so, as multiple people requested on MAPREDUCE-3760, the default active
list should only have active nodes and no unhealthy nodes.
The Java/RPC API (ClientRMProtocol#getClusterNodes) returns the list of all
active nodes, which includes unhealthy nodes. It seems confusing to me to have
them return different things. Thoughts?
bq. I don't see a point of multiple comma separated states, the default result
when no state is passed should return all active nodes. When a state is passed,
only those nodes should be removed.
My thought was that being able to specify multiple comma separated states makes
it possible to get a consistent view of the cluster. If users need to make
calls serially to get nodes in different states, they might miss some nodes
that change state. With Tucu's suggestion of an "all" option, though, maybe
this isn't necessary, so I can take it out.
> Fix up RMWebServices#getNodes
> -----------------------------
>
> Key: YARN-642
> URL: https://issues.apache.org/jira/browse/YARN-642
> Project: Hadoop YARN
> Issue Type: Bug
> Components: api, resourcemanager
> Affects Versions: 2.0.4-alpha
> Reporter: Sandy Ryza
> Assignee: Sandy Ryza
> Labels: incompatible
> Attachments: YARN-642-1.patch, YARN-642.patch
>
>
> The code behind the /nodes RM REST API is unnecessarily muddled, logs the
> same misspelled INFO message repeatedly, and does not return unhealthy nodes,
> even when asked.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
