[
https://issues.apache.org/jira/browse/HADOOP-9914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13753421#comment-13753421
]
André Kelpe commented on HADOOP-9914:
-------------------------------------
I spent a day to get the /etc/hosts and /etc/avahi/hosts right, when I created
the setup for hadoop 1.1.2 and that works fine. The namenode is actually still
working correctly: When you browse the file system, it uses the fully qualified
names of the datanodes, meaning everything works as I expect it. The
resourcemanager does not have the same same behaviour.
Here are my hosts:
$ cat /etc/hosts
127.0.0.1 localhost
192.168.7.10 master.local master
192.168.7.11 backup.local backup
192.168.7.12 hadoop1.local hadoop1
192.168.7.13 hadoop2.local hadoop2
192.168.7.14 hadoop3.local hadoop3
$ cat /etc/avahi/hosts
127.0.0.1 localhost
192.168.7.10 master.local master
192.168.7.11 backup.local backup
192.168.7.12 hadoop1.local hadoop1
192.168.7.13 hadoop2.local hadoop2
192.168.7.14 hadoop3.local hadoop3
The hostnames are also correct:
$ for host in master hadoop1 hadoop2 hadoop3; do vagrant ssh $host --command
hostname ; done
master.local
hadoop1.local
hadoop2.local
hadoop3.local
and just to be sure:
$ for host in master hadoop1 hadoop2 hadoop3; do vagrant ssh $host --command
'cat /etc/hostname' ; done
master.local
hadoop1.local
hadoop2.local
hadoop3.local
I also attached a screenshot of the webinterface.
> nodes overview should use FQDNs
> -------------------------------
>
> Key: HADOOP-9914
> URL: https://issues.apache.org/jira/browse/HADOOP-9914
> Project: Hadoop Common
> Issue Type: Bug
> Affects Versions: 2.1.0-beta
> Reporter: André Kelpe
>
> I am running a hadoop cluster in a bunch of VMs on my local machine and I am
> using avahi/zeroconf to do local name resolution (this is to avoid having to
> fiddle with my /etc/hosts file).
> The resourcemanager has an overview page, with links to all the nodemanager
> web-interfaces. The links do not work with zeroconf, due to the fact that the
> links are not including the domain part. zeroconf domains look like this
> "hadoop1.local", but the web-interface uses "hadoop1", which will not resolve.
> In hadoop 1.x all web-interfaces were using FQDN, meaning using
> avahi/zeroconf for name resolution was no problem. The same should be
> possible in hadoop 2.x.
> I am still beginning to work with hadoop 2.x, so there might be other parts,
> having the same problem, but I am not yet aware of any. If I find more of
> these, I will update this bug.
--
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
