canonical / bundle-kubeflow

Charmed Kubeflow
Apache License 2.0
102 stars 49 forks source link

The script fetches charms that are not compatible with the latest stable version of juju #153

Closed camille-rodriguez closed 4 years ago

camille-rodriguez commented 4 years ago

It seems to me that either a juju version should be highlighted during the install, or that we should not fetch charms that depends on non-stable versions of juju (2.7-rc1)

ubuntu@node07ob20:~/bundle-kubeflow$ python3 scripts/cli.py deploy-to uk8s
Enter a password to set for the Kubeflow dashboard: 
Repeat for confirmation: 
+ juju add-model kubeflow microk8s
Added 'kubeflow' model on microk8s/localhost with credential 'microk8s' for user 'admin'
+ juju deploy kubeflow --channel stable --overlay=/tmp/tmpeze0ym3g
Located bundle "cs:bundle/kubeflow-133"
Resolving charm: cs:~kubeflow-charmers/ambassador-45
Resolving charm: cs:~kubeflow-charmers/argo-controller-44
Resolving charm: cs:~kubeflow-charmers/argo-ui-45
Resolving charm: cs:~kubeflow-charmers/jupyter-controller-45
Resolving charm: cs:~kubeflow-charmers/jupyter-web-47
Resolving charm: cs:~kubeflow-charmers/katib-controller-38
Resolving charm: cs:~charmed-osm/mariadb-k8s
Resolving charm: cs:~kubeflow-charmers/katib-manager-37
Resolving charm: cs:~kubeflow-charmers/katib-ui-37
Resolving charm: cs:~kubeflow-charmers/kubeflow-dashboard-2
Resolving charm: cs:~kubeflow-charmers/kubeflow-gatekeeper-4
Resolving charm: cs:~kubeflow-charmers/kubeflow-login-4
Resolving charm: cs:~kubeflow-charmers/kubeflow-profiles-5
Resolving charm: cs:~kubeflow-charmers/metacontroller-34
Resolving charm: cs:~kubeflow-charmers/metadata-controller-2
Resolving charm: cs:~charmed-osm/mariadb-k8s
Resolving charm: cs:~kubeflow-charmers/metadata-ui-2
Resolving charm: cs:~kubeflow-charmers/minio-45
Resolving charm: cs:~kubeflow-charmers/modeldb-backend-36
Resolving charm: cs:~charmed-osm/mariadb-k8s
Resolving charm: cs:~kubeflow-charmers/modeldb-store-36
Resolving charm: cs:~kubeflow-charmers/modeldb-ui-36
Resolving charm: cs:~kubeflow-charmers/pipelines-api-45
Resolving charm: cs:~charmed-osm/mariadb-k8s
Resolving charm: cs:~kubeflow-charmers/pipelines-persistence-45
Resolving charm: cs:~kubeflow-charmers/pipelines-scheduledworkflow-46
Resolving charm: cs:~kubeflow-charmers/pipelines-ui-45
Resolving charm: cs:~kubeflow-charmers/pipelines-viewer-46
Resolving charm: cs:~kubeflow-charmers/pytorch-operator-46
Resolving charm: cs:~kubeflow-charmers/redis-36
Resolving charm: cs:~kubeflow-charmers/seldon-api-frontend-44
Resolving charm: cs:~kubeflow-charmers/seldon-cluster-manager-45
Resolving charm: cs:~kubeflow-charmers/tensorboard-44
Resolving charm: cs:~kubeflow-charmers/tf-job-dashboard-46
Resolving charm: cs:~kubeflow-charmers/tf-job-operator-44
Executing changes:
- upload charm cs:~kubeflow-charmers/ambassador-45 for series kubernetes
ERROR cannot deploy bundle: cannot add charm "cs:~kubeflow-charmers/ambassador-45": charm's min version (2.7-rc1) is higher than this juju model's version (2.6.10)
Command '('juju', 'deploy', 'kubeflow', '--channel', 'stable', '--overlay=/tmp/tmpeze0ym3g')' returned non-zero exit status 1.
knkski commented 4 years ago

Unfortunately we're relying on 2.7 feature ATM, so there aren't really any charms that work on stable Juju. Are you able to use 2.7-rc1?

camille-rodriguez commented 4 years ago

I'll switch to the latest candidate 2.7-rc4

camille-rodriguez commented 4 years ago

Maybe we could warn users in the ReadMe? The install of juju has to be specific to the channel --candidate until 2.7 is released

camille-rodriguez commented 4 years ago

even with juju-2.7rc4 I get weird stuff though. At the very end:

Deploy of bundle completed.
+ juju wait -wv
ERROR juju: "wait" is not a juju command. See "juju --help".

Did you mean:
        call
Command '('juju', 'wait', '-wv')' returned non-zero exit status 1.
knkski commented 4 years ago

Sorry about that, you'll need to run sudo snap install juju-wait --classic. It's a Juju plugin that waits for everything to get deployed properly, and to my knowledge there's no built-in support for it.

camille-rodriguez commented 4 years ago

Oh right my bad. That's in the readme haha