I would like us to try adopting the T394316: Use infrastructure as code techniques to rebuild the Beta Cluster idea as a core part of the migration. If we can setup the new hosts using OpenTofu (the FOSS fork of Terraform) we will eventually have everything moved into the new pattern.

There are some rough spots in that stack still that I have bumped into as I have tried some of this out in the past months:

• In T421244#11847209 I found that some (all?) YAML booleans in the Hiera data were being coerced to strings and breaking typing in Puppet.
• The puppetserver switch was still manual in T421244#11847986, but that may have been fixed by @dancy in T429413: Eliminate sudo rm -rf /var/lib/puppet/ssl step in new deployment-prep WMCS project (and others). Some testing needed.
• T423970: [tofu-cloudvps] Document importing legacy cloudvps_puppet_prefix objects seems likely to be solvable with some documentation on how to find a numerical prefix ID that can be used with tofu import cloudvps_puppet_prefix $ID.

Update: Southparkfan won't be available for a bit, but he and I plan to sync up in the last week of August to review things before I vanish into sabbatical and hand over stuff like this project to Tyler.

@Southparkfan contacted me yesterday with some questions and a rough outline of how to proceed with the migration. I will let them control sharing that outline and notes as seems appropriate. I do think the plan seems reasonable. Getting everything done will need some help from others, but I have confidence that @Southparkfan and @thcipriani will be able to find that help.

I have a personal hope to make some progress on T435393: Beta Cluster: Switch to PHP 8.5, includes writing Puppet changes to support PHP 8.5 on baremetal before my sabbatical with a goal of being able to show how the tofu automation can be used and more clearly define bugs in that process that were seen during a past attempt. That work is likely to set the tone for whether Southparkfan and others will be able to use tofu or not in the migration.

@bd808 has helped me to refine the plan for moving away from Buster. The plan still needs work before it can be shared and used. In short, we intend:

• to replace all Bullseye VMs with Trixie (preferred) or Bookworm (alternative) ones;
• normalise the hostnames: T280393, use deployment-docker prefix for Docker hosts, et cetera;
• codify (IaC) new VMs, their volumes, and other associated resources in Tofu.

In the meantime, some progress is being made.

1. @bd808 is working on provisioning a new deployment server through Tofu. If this succeeds, we have a boilerplate for codifying other Beta Cluster services as well.
2. Using the aforementioned boilerplate, I am trying to deploy a new Wikifeeds VM on Trixie. This VM is low-hanging fruit; not only is it "just" a Docker service definition, because we're introducing a new deployment-docker-wikifeeds prefix, we are not affected by T423970. If this succeeds, we are confident all Docker hosts - VMs with the role::beta::docker_services role - can be moved to Trixie right awey.

Traffic will work on upgrading deployment-cache-(text|upload)08 if that's fine and we would like to coordinate about that. Is this task the right place to do that or is something else preferred?

In T433998#12258671, @ssingh wrote:

Traffic will work on upgrading deployment-cache-(text|upload)08 if that's fine and we would like to coordinate about that. Is this task the right place to do that or is something else preferred?

I think a specific subtask of T401839: Migrate deployment-prep away from Debian Bullseye to Bookworm/Trixie would be the right way to track whatever work Traffic will be doing for the CDN replacement. I know that @BCornwall has started some work for that already, and I noted that in the etherpad that @Southparkfan is using for the backend planning he is doing right now.

In T433998#12258671, @ssingh wrote:

Traffic will work on upgrading deployment-cache-(text|upload)08 if that's fine and we would like to coordinate about that. Is this task the right place to do that or is something else preferred?

Of course this is fine, thank you! I have created T436468 for the Traffic team. Preferably, the new instances use a different hostname prefix *and* are codified in Tofu. Please let me know if you need assistance!