To Reproduce Steps to reproduce the behavior:

Unable to reproduce, but the workflow at the customer is during live migration.

1. Live migrate a VM with a volume attached
2. Monitor volume attachments for the volume id during and after live migration

Expected behavior

• A new volume attachment is created during the event `compute_pre_live_migration`
• When live migration completes successfully, the old volume attachment should be deleted during `compute_post_live_migration_at_destination`

Environment:

• RHOSP 17.1.5
• Ceph as storage backend
• Post copy disabled, auto converge enabled

Bug impact

• When the older volume attachments are not deleted during the live migration events, two active attachments remain in the database. This does not prevent the virtual machine or volume from working normally, so the end user doesn't notice the issue at the moment. Later, when the user performs a volume detach, only the current volume attachment is deleted, preventing the volume from being attached again somewhere else.
• Because this issue is happening on OCP worker nodes mostly, they are causing PVCs from being unavailable and wrongly identified as attached when they are not. This is disrupting OCP operations.

Known workaround

• The OCP administrators need to work with the OSP administrators to manually remove the orphan attachment.

Additional context

• Somewhere around 3k VMs were live migrated during post FFU host evacuations and it seems only 6 of them had this issue.
• The consequences of this issue were originally reported in OSPRH-18046, but we didn't understand at that moment exactly why the PVCs could not be reattached somewhere else.