[
https://issues.apache.org/jira/browse/HADOOP-11380?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14240867#comment-14240867
]
Steve Loughran commented on HADOOP-11380:
-----------------------------------------
+1 as is, though it could be good to list common problems with rack awareness
setup. You did mention missing topos, there is also
# wrong IP addresses in the script -may result in missing nodes from the
topology, even if you think you have them
# machines having wrong IP addrs. Same
# wrong machine to rack mapping: risk of scheduling "adjacent node" work on
different racks, leading to increased backbone traffic. Possibly increased
backbone traffic on writes. Possibly reduced resilience to failures.
# broken script. Check it on the CLI first. Consider keeping under SCM, maybe
even tests (e.g PyUnit)
# out of date script. Have a process for updating it. Checking out from SCM is
usually sufficient
BTW, you know you can edit documents in markdown now, under site/markdown/*.md
? Multiple editors, automatic rendering if pushed to github
> Restore Rack Awareness documentation
> ------------------------------------
>
> Key: HADOOP-11380
> URL: https://issues.apache.org/jira/browse/HADOOP-11380
> Project: Hadoop Common
> Issue Type: Bug
> Reporter: Allen Wittenauer
> Attachments: HADOOP-11380.patch
>
>
> As part of HADOOP-8427, large, extremely useful sections of the Rack
> Awareness documentation that was added in HADOOP-6616 was wiped out. We
> should restore it as a separate document.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
