JuJu stucked when I attempted to create a GPU instance on oracle cloud

[Masato Yonekawa]

In Oracle Cloud ashburn-1 region, I attempted command such as it.

$ juju add-machine --constraints "virt-type=gpu instance-type=VM.GPU3.1"

And, Juju stucked with this message.

$ juju status
Model Controller Cloud/Region Version SLA Timestamp
default oracle-cloud-controller oracle-cloud/us-ashburn-1 2.5.1 unsupported 18:07:55+09:00

Machine State DNS Inst id Series AZ Message
0 pending pending bionic failed to start machine 0 in zone "oZBk:US-ASHBURN-AD-1", retrying in 10s with new availability zone: Service error:NotAuthorizedOrNotFound. shape VM.GPU3.1 not found. http status code: 404

When I got this message, I could kill-controller only.

I could create VM.GPU3.1 instance with dashboard and oci tool.

[Tim Penhey (thumper)]

I think this is one of the general issues of removing a machine that didn't come up properly.

Did you try 'juju remove-machine 0 --force' ?

This is the documented way to remove a non-responsive machine.

[Masato Yonekawa (hyphon81)]

> Did you try 'juju remove-machine 0 --force' ?

Yes, I tried it, however I couldn't remove machine.

[Tim Penhey (thumper)]

Thanks for getting back to us.

It is likely one of the weird edge cases right now where we have something blocking the machine removal. We have someone working on these class of bugs, so we should look to get this addressed shortly.

[Anastasia (anastasia-macmood)]

I think that this is a duplicate of bug # 1814271. I'll mark it as such.

As per my comment in bug # 1814271, as of Juju 2.6, I can force remove machine in error state as well as destroy model with a machine in an error state.