Comment 0 for bug 1964124

Hi,

When deploying kubeflow on a manual central controller running 2.9.22 juju gets stuck indefinitely removing the model. trying --force, --release-storage, and deleting the namespace in Kubernetes all proved ineffective. The only workaround I could find was to login to the mongodb leader, and restart jujud. After this the model will resume removing.

We have tried leaving the model for a day or so, and the model will still fail to destroy.

Bundle exhibiting this issue: juju deploy kubeflow-lite --trust
Juju controller version: 2.9.22
model crashdump (before juju destroy-model): https://private-fileshare.canonical.com/~pjds/dsv/caas-delete-issue/juju-crashdump-4690ea05-7a61-4c25-b3f3-d7414c3cea01.tar.xz
Other notable details:

Central controller, in a different controller, manual machine (VM) controller.

Steps to reproduce:
1. Add kubeflow model: juju add-model kubeflow k8s-cloud
2. deploy kubeflow: juju deploy kubeflow-lite --trust
3. destroy model: juju destroy-model kubeflow --destroy-storage.

Accept and start the deletion.

Many thanks,

Peter