Apertium is soon going to be the last service residing on the scb cluster and needs to be migrated off of it as we want to shut down that cluster in both DCs. All services from that cluster have been or are at the last stages of being migrated to the kubernetes powered deployment pipeline.
After some discussions and exploring a variety of options , with @KartikMistry, we 've come up to the following approach that should move this forward without too many unknowns.
Move to the pipeline, using upstream Debian packages for apertium
Essentially just rely on the work that @KartikMistry has already been doing upstream (Debian) and create a pipeline image using those debian packages. That means the following:
- Use of the pipeline
- easy and orchestrated deployments/rollbacks
- No longer relying on puppet and SREs (at least to some degree) for new functionality/languages etc. When new versions of Debian packages are required, debian-backports could (and should!) be used. That decouples the work language team does from SRE.
- No dedicated hardware/VMs
- Will need to figure out if the pipeline supports this and what changes might be required.