Open davidstanke opened 1 year ago
I think there's a new doc that looks relevant here: https://cloud.google.com/workstations/docs/tutorial-automate-container-image-rebuild
More info: Per this doc, this is happening because I have a warm pool of machines... that makes sense. So we need to force a refresh. It can be done with a (IMHO janky) process of setting the pool size to zero and then back to more-than-zero, or a (IMHO better) process of using setting the container.image
field to the new container. This can probably be done via gcloud?
It appears that the machine images backing Workstations are static—they're based on a container image, but if that container changes (which can happen if the config points to the
latest
tag), the workstation config doesn't update. This means that even when the workstations are recycled, new ones aren't necessarily based on a fresh, patched container image. :(So, the pipelines for rebuilding container images (which use the
cloudbuild.yaml
configs in thecloud-workstations
folder of this repo) should be extended to also update workstation config (after the new image is built and pushed).[or, separate pipelines could monitor for new versions of the containers and then trigger workstation rebuilds based on that event]