We would like to see some increased options for the handling of secondary disks attached to VMs (ie Standard disks which are not marked as the disk's primary), particularly in the event of hardware failures.
Currently if the storage for a secondary disk cannot be reached, the VM in question essentially cannot be manipulated - Hot/Cold migrations, VM shutdowns/startups, etc all rely on onapp being able to communicate with the datastore and so will simply fail.
Additional disks attached to VMs are often not critical however, we'd therefore like to propose one or both of:
The ability to mark additional disks attached to a VM as "essential" or "non-essential", and have onapp treat them as such - So, if a VM is manipulated and onapp can't talk to an essential disk (typically an lvchange failure) it fails and rolls back as now, but if the disk is marked as non-essential onapp simply ignores the failure, doesn't attempt to mount the disk, and boots/migrates etc the VM and its remaining functioning disks.
The ability to temporarily detatch secondary disks from a VM - Effectively a database-only change to delete the associated disk configuration from a VM so that the two are no longer associated, ideally with the disk's configuration simply being moved elsehwere temporarily (rather than being dropped entirely) so that the disk can be re-attached once the faulty storage has been fixed.
The latter is essentially a process we carried out manually recently when suffering a storage failure which has spawned this request, a number of vms needed to be migrated/restarted while storage for their secondary (non-critical) disks was offline, we ended up dumping the relevant disks' configuration from the onapp database, manipulating the VMs as needed, then when the storage was brought back online, re-adding the disks' configuration.
Please sign in to leave a comment.