Changing puppet `config_version` for a running cluster does not trigger any updates on the cluster

[ocaisa]

Perhaps this is expected, but I want to transition a running cluster from 15.1.0 to 15.4.0. I thought I could do this just by updating

config_version = "15.4.0"

in main.tf, but this did not trigger any changes in Terraform. Even when I trigger a Terraform run by bringing up a node, there are no changes for the puppet repo on the management node, it is still on the 15.1.0 branch.

I think I might conceptually have the wrong idea here about how such an update should be made, but I couldn't find anything in the docs.

[cmd-ntrf]

This expected and documented here:
https://github.com/ComputeCanada/magic_castle/tree/main/docs#43-config_version

Magic Castle creates a user tf on the puppetserver and bastions. That users only has rights to update a the hieradata configuration files and the eyaml keys. It has no ability to affect the git repo containing the Puppet code itself.

The config_version is also only used during the first boot of the puppetserver to checkout the branch / revision. As we speak, Magic Castle does not have mechanism to trigger change in the Puppet code outside of connecting on the puppetserver and interacting with the git repo directly.

That said, I have come to realize yesterday while patching 14 clusters one-by-one for CopyFail, that it can be cumbersome to connect to a cluster's puppetserver to make such change. I'll start a new issue to discuss what would be the best way to do it.

[ocaisa]

Ok, thanks, matches my experience.

I too am having fun with CopyFail today.

[ocaisa]

As a side note, I did tweak puppet directly (by checking out the 15.4.0 branch), and this caused nodes with custom images to fail (a cvmfs-release was already installed so the relevant step failed as it doesn't update). It would seem that you are probably better off starting from vanilla images if you plan to bump the puppet config.