Reduces changes of things like T269825: Incorrect X-Cache-Status reported by deployment-prep caches.
We can't easily do exactly prod matching hostnames because of how hiera prefix magic is used in configuration management in Beta Cluster, but using cp instead of cache should help folks map things back to production a bit easier.
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Open | None | T53494 Use Beta cluster as a true canary for code deployments (epic) | |||
| Declined | None | T87220 Minimize infrastructure differences between Beta Cluster and production | |||
| Open | None | T401839 Migrate deployment-prep away from Debian Bullseye to Bookworm/Trixie | |||
| Resolved | BCornwall | T436468 Replace cache proxies: deployment-cache-(text|upload)08 (Bullseye deprecation) | |||
| Resolved | BCornwall | T280393 Name new CDN servers deployment-cp-(text|upload)0x instead of deployment-cache-(text|upload)0x |
I wonder at which point Traffic is planning on moving cp* hosts to bullseye? This would be much more convenient to do if there's another reason to be reimaging/renaming those hosts.
Yeah, there's no rush definitely. Just something to keep in mind for next time there's something to do around these instances.
One more issue: given cloud vps does not have per-role hiera keys, we need to rely on instance prefix based hiera (which is different for text and upload)
I think we'd have to do something like deployment-cp-(text|upload)0x exactly for this reason. I can think of a way to abuse Tofu to do things differently than just prefix puppet, but I don't think that hack would make managing the cluster easier in practice.
deployment-cp-text09 and deployment-cp-upload09 are up and running now, replacement of the incumbent hosts will be tracked in T436468