Is the RHEL 7 Packer Template for MAAS supposed to work out of the box?

jeffkight · 24 March 2020 22:50

As a test, the only change that I made was to add a --proxy <proxy server> to the kickstart file on the cloud-init repo line.
I built the image with packer, then loaded it from the command line.

Deployed a physical server with a static ip – this failed.
Deployed a physical server with a maas domain and an assigned ip on the maas controlled dhcp network – this succeeded.
Deployed a physical server with a non-dhcp network, but an assigned ip… this failed.

Initial thoughts?
Pointers for troubleshooting?

jeffkight · 25 March 2020 00:32

Note, with the mass.io provided CentOS 7 image, I have no issues with statically assigned IP addresses.

jeffkight · 25 March 2020 13:14

I noticed that the MAAS provided CentOS7 image was based on 7.6.
So I built a RHEL7.6 image… same failures.

jeffkight · 26 March 2020 00:08

Ah, after additional testing, we appear to be failing for any servers that have VLAN tagging defined.
I can install to untagged networks:

with dhcp
with auto assignment
with static assignment

That led me to check the CentOS7 image (from maas.io) a little closer. In /root/original-ks.cfg from maas.io image… it appears to be significantly different than the kickstart from the default CentOS image via packer-maas.

Will investigate the deltas and update the thread.

billwear · 26 March 2020 14:21

Thanks for your diligence, @jeffkight.

sparkiegeek · 26 March 2020 14:44

Good!

Note the packer-maas templates are not used to create the images on images.maas.io, we will be looking to unify them, but aren’t there yet.

jeffkight · 7 April 2020 22:51

So what from a cloud-init or a vlan perspective could be in play here?
Thoughts?

It would be interesting to test a RHEL 7 image made with maas-image-builder to see if VLANs are functioning there… or maybe it is a cloud-init-el issue?

ltrager · 7 April 2020 23:31

cloud-init-el is a repository that Canonical controls to provide cloud-init updates before they are added to CentOS/RHEL. This hasn’t been needed inawhile. The version in that repository is currently older then what CentOS/RHEL ship. So the installer should be using the version from CentOS/RHEL.

When the failure occurs can you confirm that the system boots into RHEL? If it doesn’t you can boot into rescue mode, mount the filesystem and the systems logs for errors.