Add TX2 cluster update d05 cluster and preseed naming

[BaptisteGerondeau]

• Adds TX2 cluster and associated preseeds
• Updates the d05 cluster with the new machines
• Switches d05 cluster to use Centos-Upstream
• Each machine type has its associated preseed for Debian ERP and CentOS
• Adds TX2 cluster to ohpc jobs
• Adds variables to ohpc_install for IB installation

Also, a consequence of this is that ohpc_install can easily be edited to work in a "degraded" mode, where you can exclude certain machines from the cluster.

[rengolin]

Why do we need one preseed/kickstart/autoyast for every type of machine?

This will encourage updates in one but not others and will fragment the lab.

We only need 5:

• CentOS ERP + Upstream
• Debian ERP + Upstream
• OpenSUSE Upstream

We have yet no reason to do a Suse ERP, as it should work upstream on all three clusters (same as CentOS).

We only need CentOS ERP for now to make sure it doesn't get discarded again and we need it in future hardware.

We use Debian ERP because of the D03 and the need to use the same system on all benchmark machines, but we should also eventually test Debian upstream (not for this patch, though).

[BaptisteGerondeau]

I do agree that this system does introduce a lot of redundant code, and it would work better in a template system (which is outside of the scope of this PR) which would eliminate most of the redundancy (at the cost of increased complexity).

I'll roll back this change and force push.

[BaptisteGerondeau]

@rengolin

[rengolin]

The main change from TX2 and others is the swap, right? Did you try that on the other clusters?

I think we ought to use no swap on all clusters and have a single preseed for all.

[BaptisteGerondeau]

All machines using preseeds/kickstarts with manual partitioning to disable swap, thus the number of preseeds/kickstarts is reduced to only 3.