[
https://issues.apache.org/jira/browse/HADOOP-2776?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Allen Wittenauer resolved HADOOP-2776.
--------------------------------------
Resolution: Won't Fix
I'm going to close this as won't fix. I don't think this is anything that we
actually can fix here other than providing a complicated hostname mapping
system for web interfaces.
Part of the frustration I'm sure stems from a misunderstanding of what is
actually happening:
bq. The slaves file has the public names listed.
The slaves file is only used by the shell code to run ssh connections. It has
absolutely zero impact on the core of Hadoop.
bq. Resolving a public name inside EC2 returns the private IP (which would
reverse to the internal DNS name).
Hadoop makes the perfectly valid assumption that the hostname the system tells
us is a valid, network-connectable hostname. It is, from the inside of EC2.
We have no way to know that you are attempting to connect from a completely
different address that is being forwarded from some external entity.
Proxying connections into a private network space is a perfectly valid
solution.
> Web interface uses internal hostnames on EC2
> --------------------------------------------
>
> Key: HADOOP-2776
> URL: https://issues.apache.org/jira/browse/HADOOP-2776
> Project: Hadoop Common
> Issue Type: Bug
> Components: contrib/cloud
> Affects Versions: 0.15.1
> Environment: EC2 ami-a324c1ca
> Reporter: David Phillips
>
> The web interface, for example http://$MASTER_HOST:50030/machines.jsp, uses
> internal hostnames when running on EC2. This makes it impossible to access
> from outside EC2.
> The slaves file has the public names listed. Resolving a public name inside
> EC2 returns the private IP (which would reverse to the internal DNS name).
--
This message was sent by Atlassian JIRA
(v6.2#6252)
