[
https://issues.apache.org/jira/browse/YARN-4517?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15193451#comment-15193451
]
Varun Saxena commented on YARN-4517:
------------------------------------
Thanks [~gtCarrera9] for the review.
bq. One main question: maybe we want to unify all application/container views?
I noticed that right now, the "application views" from the application list and
from the NM are different. Ideally, we'd like to provide one unified place to
show one application, no matter the user arrives from the app list, flow list,
NM app list or anywhere else? We can also integrate Similar story also applies
to the container view?
The application and container states in NM will be distinct from RM.
The applications and containers seen here are running containers(applications
are seen a bit longer based on keep alive time - depends on config). This
information including NM states by itself can be useful.
However, I think we can fit some container related information(which is
fetched from NM) on the main container page. We can leave out some unnecessary
info too. This is more a mimic of what was there on old UI.
Will have to discuss page layouts and organization in detail.
bq. Meanwhile, maybe it's time to start detailed page style designs. With
unified app/container views, we need to address questions like where to put
node id/app ids in the page, and how to organize all available data on the page?
Sure. Suggestions are welcome. I agree we need to discuss this in detail and
reach a consensus on what goes where. Even I have not given a great deal of
thought on this. Maybe after we move this branch's code into trunk. Because
this is important to go in for YARN-2928.
bq. I noticed one workflow related problem: once a NM is in shutdown state, it
is not possible to go into the node page. What is the assumed debug workflow on
this?
The link to node page has been disabled because we cannot reach NM in this
state. What information are you expecting to give here ? A way to display NM
logs, for instance ?
bq. On my local machine, links to application logs are broken with just a 500
error. Maybe we can improve this in future.
Where ? On the app page ?
bq. Seems like there's no need to show "node labels" if node label is not
enabled?
In UI, we cannot know if node labels are enabled or not(until and unless we
iterate over the whole output and assume if labels are not there for any node,
it means labels are not enabled). Even if labels are enabled but not attached
to any node, output will be same. Maybe once we start ember from within
RM(there is a JIRA for this), we can think about using these configurations.
Thoughts ?
bq. I'm not sure about the meaning of the row in node status showing "Node
Health Report".
In NM, we can configure disk health check(true by default) and health check
scripts. Node Health report will contain output from that. It will contain
information about which disks are bad, for instance. In normal case, it will be
empty.
> [YARN-3368] Add nodes page
> --------------------------
>
> Key: YARN-4517
> URL: https://issues.apache.org/jira/browse/YARN-4517
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: yarn
> Reporter: Wangda Tan
> Assignee: Varun Saxena
> Labels: webui
> Attachments: (21-Feb-2016)yarn-ui-screenshots.zip,
> Screenshot_after_4709.png, Screenshot_after_4709_1.png,
> YARN-4517-YARN-3368.01.patch, YARN-4517-YARN-3368.02.patch
>
>
> We need nodes page added to next generation web UI, similar to existing
> RM/nodes page.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
