Resize does not allow you to resize down in mitaka

[Kevin]

Deployed cluster with Fuel 9.0.

Created and launched an instance with resources of 2VCPUs, 8GB of ram, 100GB ephemeral storage. Ephemeral storage is not backed by CEPH.

I was able to successfully use Horizon to resize the instance (CentOS) from the current running resources above to a larger instances with 4VCPUS, 16GB of ram and 200GB of ephemeral storage. However, when I went to go back down to the 2VCPUS, 8GB Ram, 100GB ephemeral the instance won't resize.

In horizon

1. click on the "Resize" option
2. Confirm the resize
3. Horizon displays a "Success" message (but actually failed in the background)

I check the instances "Action Log" in horizon and I see that it errored. Go check the logs on the controller and see.

WARNING nova.scheduler.host_manager [req-d46ad3a1-be18-464b-a3b9-11123e481fcc bdb162ee567d4230a988895f2a000a8b9 84ec9bb0ccc34eea84fbf49b557c4a66] Host has more disk space than database expected (43gb > 34gb)

Searching around on the internet I found this https://ask.openstack.org/en/question/43359/resize-openstack-icehouse-instance-at-same-node-bug/ and it isn't clear that this was a bug or not. Some people have reported that it works for them in older versions but not in newer version. Others report that this is a problem resizing down in that it should just show an error message. I just need clarification if this is a bug, it is intended to error on a resize down (granted I don't think it should show success in then failure).

[Maciej Szankin (mszankin)]

This is not a bug, in my opinion.

Resizing down works when two flavors have the same disk params. Resizing down disk is not permitted due to creating potential filesystem fault.

[Kevin (kvasko)]

This might need to be transferred over to the Horizon project then but if that is the case there should be an error message back to the user instead of a green "success" when you confirm the resize.