ResourceManager not matching host names

[Roger Hoover <ro...@gmail.com>]

Hi,

I'm having a issue with matching hostnames in the scheduler for slave nodes
whose hostnames do not match their fully qualified domain names.

The problem is that the ResourceManager learns the hostname of the node
when the NodeManager registers itself and it seems the node manager is
getting the hostname by asking the OS.  When a job is submitted, I think
the ApplicationMaster learns the hostname by doing a reverse DNS lookup
based on the slaves file.

Therefore, the ApplicationMaster submits requests for containers using the
fully qualified domain name (foo.bar.com) but the scheduler uses the OS
hostname (foo) when checking to see if any requests are node-local.

What's the recommended solution?
a) Always make sure you configure you clusters such that hostnames match
reverse DNS?
b) Create a way for the ApplicationMaster and NodeManager to agree on
hostnames?

Thanks,

Roger

Re: ResourceManager not matching host names

[Roger Hoover <ro...@gmail.com>]

Bikas,

Thank you for your response.  I've submitted a JIRA ticket here with steps
to reproduce (https://issues.apache.org/jira/browse/YARN-466).
 Unfortunately, I don't have bandwidth to work on a patch.

Cheers,

Roger


On Mon, Mar 11, 2013 at 10:58 AM, Bikas Saha <bi...@hortonworks.com> wrote:

> Thanks for the investigation. Could you please open a JIRA to track this.
> Would be great if you can add repro steps. Of course, patch to fix is most
> welcome.
>
> Bikas
>
> -----Original Message-----
> From: Roger Hoover [mailto:roger.hoover@gmail.com]
> Sent: Friday, March 08, 2013 5:12 PM
> To: yarn-dev@hadoop.apache.org
> Subject: ResourceManager not matching host names
>
> Hi,
>
> I'm having a issue with matching hostnames in the scheduler for slave
> nodes whose hostnames do not match their fully qualified domain names.
>
> The problem is that the ResourceManager learns the hostname of the node
> when the NodeManager registers itself and it seems the node manager is
> getting the hostname by asking the OS.  When a job is submitted, I think
> the ApplicationMaster learns the hostname by doing a reverse DNS lookup
> based on the slaves file.
>
> Therefore, the ApplicationMaster submits requests for containers using the
> fully qualified domain name (foo.bar.com) but the scheduler uses the OS
> hostname (foo) when checking to see if any requests are node-local.
>
> What's the recommended solution?
> a) Always make sure you configure you clusters such that hostnames match
> reverse DNS?
> b) Create a way for the ApplicationMaster and NodeManager to agree on
> hostnames?
>
> Thanks,
>
> Roger
>

RE: ResourceManager not matching host names

[Bikas Saha <bi...@hortonworks.com>]

Thanks for the investigation. Could you please open a JIRA to track this.
Would be great if you can add repro steps. Of course, patch to fix is most
welcome.

Bikas

-----Original Message-----
From: Roger Hoover [mailto:roger.hoover@gmail.com]
Sent: Friday, March 08, 2013 5:12 PM
To: yarn-dev@hadoop.apache.org
Subject: ResourceManager not matching host names

Hi,

I'm having a issue with matching hostnames in the scheduler for slave
nodes whose hostnames do not match their fully qualified domain names.

The problem is that the ResourceManager learns the hostname of the node
when the NodeManager registers itself and it seems the node manager is
getting the hostname by asking the OS.  When a job is submitted, I think
the ApplicationMaster learns the hostname by doing a reverse DNS lookup
based on the slaves file.

Therefore, the ApplicationMaster submits requests for containers using the
fully qualified domain name (foo.bar.com) but the scheduler uses the OS
hostname (foo) when checking to see if any requests are node-local.

What's the recommended solution?
a) Always make sure you configure you clusters such that hostnames match
reverse DNS?
b) Create a way for the ApplicationMaster and NodeManager to agree on
hostnames?

Thanks,

Roger