Interface mtu property

[jhoblitt]

I use jumbo frames on a few subnets so I've taken a stab at adding a mtu property to the network_config type and corresponding support in the redhat provider. This has only been tested on SL6.4. I haven't yet tried to implement debian family provider support but am happy to take that on before this PR is merged if this type API change is potentially mergable.

# puppet resource --modulepath=`pwd` network_config eth0
network_config { 'eth0':
  ensure    => 'present',
  family    => 'inet',
  ipaddress => '10.10.10.10',
  method    => 'static',
  mtu       => '9000',
  netmask   => '255.255.255.0',
  onboot    => 'true',
  options   => {'HWADDR' => '00:1F:C6:16:0A:91', 'IPV6INIT' => 'no', 'NM_CONTROLLED' => 'no', 'TYPE' => 'Ethernet'},
}

[ericsakowski]

@jhoblitt - I'm confused by the need for this, why not just pass options => {'MTU' => '9000'} ?

[jhoblitt]

@ericsakowski Setting the MTU on an interface is a property that should be universally abstractable for all platforms. Keys passed to options are not validated and are inherently platform specific. By your logic, why not remove all the properties from the the type except options?

[ericsakowski]

@jhoblitt Good point -- sometimes I forget there are other platforms than RHEL.

[adrienthebo]

It's generally preferred to use interpolation over format strings, EG "#{value} is not an valid mtu [...]

[adrienthebo]

And tiny grammatical nitpick: it's better english to say #{value} is not a valid mtu :)

[adrienthebo]

I'm 👍 on this, but I would like to add Debian support to this before it's merged, or add this as a feature.

[jhoblitt]

ACK all those requests. I was hoping to get to the debian provider
yesterday but I'm currently in a $day_job high priority interrupt. :)

[jhoblitt]

@adrienthebo I believe I've addressed your comments, made a few other minor tweaks and taken a stab the debian provider. The later still needs testing on real world boxen. It's the end of the day so I pushed but I can't keep github from updating the PR. I also rebased this branch on the current master.

I'm trying to decide if parsed mtu values should be checked inside of the provider or if trying to validate parsed values is the road to madness...

[jhoblitt]

I've finally had some time to manually test this on a debian squeeze VM. I think it's ready for another review and/or merging.

[adrienthebo]

Typo: 'postive' :)

[adrienthebo]

@jhoblitt looks great! I had two tiny nitpicks, but once those are addressed I think this is good. :)

[jhoblitt]

@adrienthebo I've fixed those two legitimate issues. ;) I think the patch set is in a mergable state now.

[adrienthebo]

BRILLIANT!

Merged into master in 53a3d14; this should be released in 0.5.0, thanks again! Awesome work!