[
https://issues.apache.org/jira/browse/YARN-3040?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14378666#comment-14378666
]
Zhijie Shen commented on YARN-3040:
-----------------------------------
Thanks for review, Sangjin and Junping! I've updated the patch accordingly.
bq. I am comfortable with continuing to work on the flow-related items in the
separate JIRA.
Thanks. This sounds good.
bq. I'm not sure of these set calls. Are these here just to initialize the
context to default values?
Yes, these are the defaults. In fact, "user" will sure to be update by the rpc
call to get the context info (unless there's bug in the RPC). The current user
for initialization is usually not correct, but I kept to have a value to ensure
we have something to pass to the storage to prevent possible NPE that will
crush the process. Instead, we can easily debug/inspect the storage to verify
the user if bug occurs. I add some code comments for the initialization.
bq. I would prefer something like "yarn.cluster.id" because this id is for
identifying YARN cluster rather than ResourceManager.
I also agree "yarn.cluster.id" sounds better, but
"yarn.resourcemanager.cluster-id" is the legacy name, which is used by RM HA
for a while. As it's not sound so bad, how about keeping it, such that we don't
need to deprecate config or break compatibility.
bq. Can we add a test case that without specifying flow_id and flow_run_id and
v2 timeline service still can work?
Added the test case in the new patch
bq. Do we need to be case-insensitive here? I think we can be strict about the
tag names?
This is because the tag text has case sensitive and insensitive mode. When
insensitive, even if user inputs the upper case strings, it will be normalized
to lower case strings. So we need to take care this case.
bq. You might want to be bit defensive about the tag not carrying any value
(e.g. "TIMELINE_FLOW_ID_TAG:").
It will check if the tag starts with "TIMELINE_FLOW_ID_TAG:", and then if the
value is empty,
{{"TIMELINE_FLOW_ID_TAG:".substring("TIMELINE_FLOW_ID_TAG".length() + 1)}} will
return an empty value. It shouldn't throw IndexOutOfBoundsException. But it
seems there's no need to add an empty env, I'll change the code accordingly.
In addition, I fixed a couple test failure in the new patch.
> [Data Model] Make putEntities operation be aware of the app's context
> ---------------------------------------------------------------------
>
> Key: YARN-3040
> URL: https://issues.apache.org/jira/browse/YARN-3040
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: timelineserver
> Reporter: Sangjin Lee
> Assignee: Zhijie Shen
> Attachments: YARN-3040.1.patch, YARN-3040.2.patch, YARN-3040.3.patch
>
>
> Per design in YARN-2928, implement client-side API for handling *flows*.
> Frameworks should be able to define and pass in all attributes of flows and
> flow runs to YARN, and they should be passed into ATS writers.
> YARN tags were discussed as a way to handle this piece of information.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
