server cannot reboot after multi-attach volume retyping

[zhaoleilc]

Description
===========
The server will fail to hard reboot after an retype operation
is performed on the multi-attach volume of it.
With multiattach the nova will call to volume api to get volume
according to connection_info in block device mappings of nova.
However, the volume id may become name_id after retyping
operation and server will fail to hard reboot since volume
cannot be find.

Expected result
===============
The server will succedd in hard reboot after an retype operation
is performed on the multi-attach volume of it.

The server will fail to hard reboot after an retype operation
is performed on the multi-attach volume of it.

[Artom Lifshitz (notartom)]

The first thing that I notice when I look at the Cinder api-ref [1] is that they discourage retyping an in-user multiattach volume if there are multiple read-write attachments:

"Retyping an in-use volume from a multiattach-capable type to a non-multiattach-capable type, or vice-versa, is not supported. It is generally not recommended to retype an in-use multiattach volume if that volume has more than one active read/write attachment."

That being said, could you provide the nova-compute log with the error that happens when you attempt to reboot an instance with a volume that's been retyped?

[1] https://docs.openstack.org/api-ref/block-storage/v3/index.html?expanded=retype-a-volume-detail#retype-a-volume

[Launchpad Janitor (janitor)]

[Expired for OpenStack Compute (nova) because there has been no activity for 60 days.]