Will rancher-vcluster work air-gapped?

[Martin-Weiss]

What will happen if the repos in this yaml are not reachable due to air-gapped?

experimental-addons/rancher-vcluster/rancher-vcluster.yaml

Line 15 in f0dccb5

repo: https://charts.harvesterhci.io

[tserong]

I gave this a quick test by setting this to a URL I know isn't reachable. It will fail after a few minutes:

If you happen to check the logs while it's trying to apply, you'll see something helpful:

# kubectl -n rancher-vcluster logs job/helm-install-rancher-vcluster
[ ... ]
+ helm repo add rancher-vcluster https://192.168.0.44
Error: looks like "https://192.168.0.44" is not a valid chart repository or cannot be reached: Get "https://192.168.0.44/index.yaml": dial tcp 192.168.0.44:443: connect: no route to host

If you don't think to do that, AFAICT the logs will be gone and you'll just have this with no real useful information about what's wrong:

# kubectl -n rancher-vcluster describe jobs/helm-install-rancher-vcluster | tail
[...]
Events:
  Type     Reason                Age    From            Message
  ----     ------                ----   ----            -------
  Normal   SuccessfulCreate      8m58s  job-controller  Created pod: helm-install-rancher-vcluster-p86tc
  Normal   SuccessfulDelete      5m40s  job-controller  Deleted pod: helm-install-rancher-vcluster-p86tc
  Warning  BackoffLimitExceeded  5m40s  job-controller  Job has reached the specified backoff limit

If the various URLs in rancher-vcluster.yaml were pointed to local registries inside the airgap (i.e. Edit the YAML in the Harvester UI), it should work fine.

[Martin-Weiss]

I guess we need to deliver this helm chart and all yaml as part of our harvester delivery and not rely on sources from the internet - as otherwise the process does not work air-gapped... and we need to add the list of required images to the images.txt and iso?