geni-get manifest failure

[Jeanne Ohren]

Gary,

I'm just getting myself familiar with the geni-get tool. I am finding
that the
manifest command does not work for me on my instaGENI nodes.
I have tried this on both OpenVz and Xen-Ubuntu12.04 nodes.
Is this command not yet supported?

On Xen (Ubuntu12.04): [Wisconsin IG]
vm-1:~% geni-get manifest
/usr/bin/geni-get: manifest: internal error handling request
vm-1:~% geni-get --version
1
vm-1:~% geni-get --client-version
1.1

On OpenVz: [Clemson IG]
[johren@router /tmp]$ geni-get manifest
internal error handling request
[johren@router /tmp]$ geni-get --version
1
[johren@router /tmp]$ geni-get --client-version
1

Thanks,
--Jeanne

--
Jeanne Ohren
Software Engineer
Raytheon BBN Technologies
Office: 952.545.5726
Cell: 612.963.0395
Jabber: joh...@j.ir.bbn.com

[Gary Wong]

On Tue, Apr 01, 2014 at 02:36:23PM -0500, Jeanne Ohren wrote:
> I'm just getting myself familiar with the geni-get tool. I am finding
> that the
> manifest command does not work for me on my instaGENI nodes.
> I have tried this on both OpenVz and Xen-Ubuntu12.04 nodes.
> Is this command not yet supported?

Hi,

It is supposed to be supported.

> On Xen (Ubuntu12.04): [Wisconsin IG]
> vm-1:~% geni-get manifest
> /usr/bin/geni-get: manifest: internal error handling request
> vm-1:~% geni-get --version
> 1
> vm-1:~% geni-get --client-version
> 1.1

I'm afraid I can't reproduce the problem. I brought up a sliver on
the Wisconsin InstaGENI rack (using the attached rspec -- a single
Ubuntu 12.04 node in a Xen guest), and it works for me:

node:~% geni-get manifest
<rspec xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns="http://www.geni.net/resources/rspec/3" type="manifest" xsi:schemaLocation="http://www.geni.net/resources/rspec/3 http://www.geni.net/resources/rspec/3/manifest.xsd">
<node client_id="node" exclusive="false" component_id="urn:publicid:IDN+instageni.wisc.edu+node+pc3" component_manager_id="urn:publicid:IDN+instageni.wisc.edu+authority+cm" sliver_id="urn:publicid:IDN+instageni.wisc.edu+sliver+4174">
<sliver_type name="emulab-xen">
<disk_image name="urn:publicid:IDN+emulab.net+image+emulab-ops//UBUNTU12-64-STD"/>
</sliver_type>
<rs:vnode xmlns:rs="http://www.protogeni.net/resources/rspec/ext/emulab/1" name="pcvm3-5"/><host name="node.gtwtest.emulab-net.instageni.wisc.edu"/><services><login authentication="ssh-keys" hostname="pc3.instageni.wisc.edu" port="31802" username="gtw"/></services></node>
node:~% geni-get --version
1
node:~% geni-get --client-version
1.1

Could you please try creating another sliver at Wisconsin using this
rspec, and let me know if it works? If it fails, please leave the sliver
active, and I'll inspect further.

Thanks,
Gary.
--
Gary Wong g...@flux.utah.edu http://www.cs.utah.edu/~gtw/

[Jeanne Ohren]

Thanks, Gary. I created a sliver on Wisc IG and the geni-get manifest
command
does work for me on that sliver (joGeniGet).

This made me think that it might be an issue with either the image that
I was using
or the racks that I was using so I tried a sliver starting with your
rspec and adding
another node with the image I was using previously. I found in that case
that both nodes gave me the "internal error handling request" error.
So I then tried an image using your rspec at wisconsin only I duplicated
the one
node to create two identical nodes. Again, both nodes give this error.

Could it be an issue with having more than one node in the sliver?

I still have both slivers up at wisconsin if you want to take a look:

joGeniGet: created with your rspec (one node)
joGeniGet2: starting with your rspec and adding a second identical node

--Jeanne

[Gary Wong]

On Thu, Apr 03, 2014 at 02:07:17PM -0500, Jeanne Ohren wrote:
> Could it be an issue with having more than one node in the sliver?
>
> I still have both slivers up at wisconsin if you want to take a look:
>
> joGeniGet: created with your rspec (one node)
> joGeniGet2: starting with your rspec and adding a second identical node

Thanks very much -- having the two to compare turned out to be very
useful. You are on the right track with it being related to more than
one node... it is not exactly that, but rather related to the size of the
rspec (which will of course grow bigger as nodes are added). Flack
makes things worse, since it includes an awful lot of information in
the rspecs it generates (everything from the screen resolution of the
client to the undo history).

I'll test a fix locally, and if all goes well then hopefully
"geni-get manifest" will work with big rspecs on InstaGENI racks after
tomorrow's maintenance window.

[Jeanne Ohren]

Ah, yes. That makes sense. Excellent! I'll check again after the
window tomorrow.

Thanks, Gary!
--Jeanne

[Gary Wong]

On Thu, Apr 03, 2014 at 05:49:20PM -0500, Jeanne Ohren wrote:
> Ah, yes. That makes sense. Excellent! I'll check again after the
> window tomorrow.

Thanks! If this is blocking something for you, this fix is installed
now on emulab.net. But if you can wait a day, it should be available
on all InstaGENI racks in around 24 hours.

Bye for now,