Wrong PXE interface name during provisioning

[Aviram Bar-Haim]

At the end of the provisioning stage (using image based provisioning), the installation got stuck.
When looking into nodes I figured out that my PXE interface is defined as eth2 in udev but only ifcfg-eth0 exists in /etc/sysconfig/network-scripts.

after accessing iva ILO and changing eth0 configurations to eth2 and restart network service it worked again.

{"build_id": "2015-04-04_05-40-09", "ostf_sha": "b978badb43243d3f1b85dde15b8e301a21f12254", "build_number": "155", "release_versions": {"2014.2-6.1": {"VERSION": {"build_id": "2015-04-04_05-40-09", "ostf_sha": "b978badb43243d3f1b85dde15b8e301a21f12254", "build_number": "155", "api": "1.0", "nailgun_sha": "a2dc04af9c5838a990adeab81789d4c090943701", "production": "docker", "python-fuelclient_sha": "9ad42671130deb2cc81135fb453ef0793883be2c", "astute_sha": "240bfefe0a0b5af21b79854d0c96d8a3ee0fd885", "feature_groups": ["experimental"], "release": "6.1", "fuelmain_sha": "f90d7d40ceff39ee6d7587b183c63bd00ed0743f", "fuellib_sha": "601cc8e1dccbdcf328f83f071270609e824d6c5a"}}}, "auth_required": true, "api": "1.0", "nailgun_sha": "a2dc04af9c5838a990adeab81789d4c090943701", "production": "docker", "python-fuelclient_sha": "9ad42671130deb2cc81135fb453ef0793883be2c", "astute_sha": "240bfefe0a0b5af21b79854d0c96d8a3ee0fd885", "feature_groups": ["experimental"], "release": "6.1", "fuelmain_sha": "f90d7d40ceff39ee6d7587b183c63bd00ed0743f", "fuellib_sha": "601cc8e1dccbdcf328f83f071270609e824d6c5a"}

Diag snapshot: https://drive.google.com/file/d/0B26TG8UdpL-ESzd0RmtrWUFzSFk/view?usp=sharing

[Vladimir Kozhukalov (kozhukalov)]

this bug has been fixed https://review.openstack.org/#/c/169736/

[Alexander Gordeev (a-gordeev)]

I don't think it's a duplicate.

I've double checked (by date and by sha) nailgun which was shipped inside of ISO already includes this fix https://review.openstack.org/#/c/169736/

[Andrey Danin (gcon-monolake)]

cloud-init.log from node-3

[Vladimir Kozhukalov (kozhukalov)]

Still tend to think that is a duplicate.

Aviram, could you please clarify whether you touched bootstrap image /var/www/nailgun/bootstrap/initramfs.img.gz or not before starting deployment. Im asking because the data that were passed to fuel agent were correct (eth2 is to be admin interface) and this bug (centos + IBP + wrong admin interface) has definitely been fixed in the version of fuel that you used.

If you need to modify bootstrap initramfs somehow, please make sure you modify the version with this fix included https://review.openstack.org/#/c/169736/ .

[Aviram Bar-Haim (aviramb)]

Vladimir, We're changing the /var/www/nailgun/bootstrap/initramfs.img before starting the deployment in our plugin installation.
Does IBP relyes on a specific bootstrap image? if so, should we rebuild bootstrap image with the current fuel-main master?

[Vladimir Kozhukalov (kozhukalov)]

Aviram,

IBP uses Fuel Agent https://github.com/stackforge/fuel-web/tree/master/fuel_agent. This is the piece of software which downloads OS images from the master node and puts them on hard drives. It also prepares some data for cloud-init to do initial configuration (including configuring admin interface) after first reboot. We install the corresponding RPM package into the bootstrap ramdisk. The issue with configuring admin interfaces was the Fuel Agent issue and it has been fixed. So, if you build your own bootstrap ramdisk, you definitely need to install Fuel Agent with all recent fixes. To be more particular, this bug about admin interfaces has been fixed by this patch https://review.openstack.org/#/c/169736/