ErrImagePull on k8s deploys with freshly bootstrapped 2.8-rc1 controller

[Paul Collins]

I bootstrapped a controller in LXD with the current beta snap:

$ juju version
2.8-rc1-focal-amd64
$ snap info juju | grep ^installed:
installed: 2.8-rc1 (11992) 134MB classic
$ snap refresh juju
snap "juju" has no updates available
$ _

Deploying k8s charms to the LXD controller fails with the following error:

  Warning Failed 2m58s (x3 over 3m49s) kubelet, agnew Failed to pull image "jujusolutions/jujud-operator:2.8-rc1.3638": rpc error: code = Unknown desc = failed to resolve image "docker.io/jujusolutions/jujud-operator:2.8-rc1.3638": no available registry endpoint: docker.io/jujusolutions/jujud-operator:2.8-rc1.3638 not found

I bootstrapped a k8s controller with this same client, and it worked, but seems to be using a different image:

    Image: jujusolutions/jujud-operator:2.8-rc1.3639

However, doing a juju deploy to this controller fails in the same way:

  Warning Failed 13s kubelet, agnew Failed to pull image "jujusolutions/jujud-operator:2.8-rc1.3638": rpc error: code = Unknown desc = failed to resolve image "docker.io/jujusolutions/jujud-operator:2.8-rc1.3638": no available registry endpoint: docker.io/jujusolutions/jujud-operator:2.8-rc1.3638 not found

[Ian Booth (wallyworld)]

There was an issue with the non-stable release build scripts which added a build number by mistake to the juju binary.

I've added a tag to the operator image so get around the issue until rc2. So it should work now.

[Ian Booth (wallyworld)]

The fix is to the build scripts which are maintained outside juju.
Since rc1 now can bootstrap and deploy k8s charms, I'll mark as fix committed as far as juju is concerned.