[
https://issues.apache.org/jira/browse/HIVE-15971?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15876814#comment-15876814
]
Siddharth Seth commented on HIVE-15971:
---------------------------------------
+1. Looks good. There's a couple of follow up items.
Reading the port from YARNConfiguration is not great. A port value of 0 means a
dynamic port, in which case this completely breaks. We need a follow up jira to
figure out a good way to make the port available. Likely published from within
the LLAPDaemon container itself.
Other than this, if the llap instance is not found (e.g. task timed out because
llap went down) - we won't be able to construct the log URL. Probably need to
handle this by retaining information for some time.
A bunch of this can be simplified if NodeId could be extended in Yarn. A
LlapNodeId could include information about the container, NM webaddress, etc -
at allocation time.
> LLAP: logs urls should use daemon container id instead of fake container id
> ---------------------------------------------------------------------------
>
> Key: HIVE-15971
> URL: https://issues.apache.org/jira/browse/HIVE-15971
> Project: Hive
> Issue Type: Bug
> Components: llap
> Affects Versions: 2.2.0
> Reporter: Prasanth Jayachandran
> Assignee: Prasanth Jayachandran
> Attachments: HIVE-15971.1.patch, HIVE-15971.2.patch,
> HIVE-15971.3.patch, HIVE-15971.4.patch
>
>
> The containerId used for log url generation is fake. It should be replaced by
> the container id of the llap daemon.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
