docs: describe how slave nodes to choose which interface is used for PXE booting

[Andrew Woodward]

We received bp: https://blueprints.launchpad.net/fuel/+spec/slave-nodes-change-pxe-iface

> It would be helpful if slave nodes have the ability to pick the interface used for PXE bootstrapping. Some users are having trouble that eth0 is a more powerful NIC (10g) than is needed for PXE bootstrapping.

however, the nic is automatically selected based on the receiving network.

We need to document
a) how this works
b) how to troubleshoot it

probably both as a LP answer as well as in the OPS guide

----

Fuel configures the nic named/order based on the data seen by the nailgun agent (/opt/nailgun/bin/agent) on discovered nodes. This is in turn the result of how the nics are named/ordered on by the boot strap node.

The device used by the admin network will be the interface that is directly attached to the Admin PXE network, if one is not available it will fall back to the interface with the default gateway.

You may have

PHYS DEV MAC
0 eth0 :FE:A0
1 eth1 :BC:6D
2 eth2 :E1:B2

If physical device 0 is connected to the Admin PXE network, then eth0 will be the admin interface in fuel

If instead physical device 1 is connected to the Admin PXE network, then eth1 will be the admin interface in fuel.

A common issue here is that physical device 0, may not always be assigned device eth0. You may see:

PHYS DEV MAC
0 eth2 :FE:A0
1 eth0 :BC:6D
2 eth1 :E1:B2

In this case, having physical device 0 connected to the Admin PXE network will result in the eth2 interface being used as the admin interface in fuel

You can confirm that the right interface is in use because the MAC address didn't change even though the device name did.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/153153

[Ryan Moe (rmoe)]

Medium priority bugs are not backported to stable releases.

[Ryan Moe (rmoe)]

Sorry for the noise, I missed that this was customer found.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/153153
Committed: https://git.openstack.org/cgit/stackforge/fuel-docs/commit/?id=3bb50ef5a2d0e0bce27780e93955bc975cf49144
Submitter: Jenkins
Branch: master

commit 3bb50ef5a2d0e0bce27780e93955bc975cf49144
Author: Irina Povolotskaya <email address hidden>
Date: Thu Feb 5 10:26:57 2015 +0300

    Describe how slave nodes choose the interface when PXE booting

    This change contains the logic of PXE booting for slave nodes
    and explains the situation with eth interfaces and physical device
    numbers.

    Change-Id: Ia9ce80cfff1718c2942263fca2ca68bb41e458b9
    Closes-Bug: 1417801

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/6.0
Review: https://review.openstack.org/165133

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/165133
Committed: https://git.openstack.org/cgit/stackforge/fuel-docs/commit/?id=4d6d4ce0756bee3d4a8218d8e75892b6f96ce7f0
Submitter: Jenkins
Branch: stable/6.0

commit 4d6d4ce0756bee3d4a8218d8e75892b6f96ce7f0
Author: Irina Povolotskaya <email address hidden>
Date: Thu Feb 5 10:26:57 2015 +0300

    Describe how slave nodes choose the interface when PXE booting

    This change contains the logic of PXE booting for slave nodes
    and explains the situation with eth interfaces and physical device
    numbers.

    Change-Id: Ia9ce80cfff1718c2942263fca2ca68bb41e458b9
    Closes-Bug: 1417801
    (cherry picked from commit 3bb50ef5a2d0e0bce27780e93955bc975cf49144)