| 2023-08-08 12:05:44 |
Dimitri John Ledkov |
description |
Currently qemu & openstack for the most simple invocations of "virtio" drive type use virtio-blk.
One can use virtio-scsi, and many do, but it's not a given.
To continue support minimal images without initrd, and enable in-place no-change migration, we should also make virtio-blk builtin.
This can be revisited once openstack import defaults to "--property hw_disk_bus=scsi --property hw_scsi_model=virtio-scsi", and similarly qemu default virt model depends on virtio-scsi as well. |
Currently qemu & openstack for the most simple invocations of "virtio" drive type use virtio-blk.
One can use virtio-scsi, and many do, but it's not a given.
To continue support minimal images without initrd, and enable in-place no-change migration, we should also make virtio-blk builtin.
This can be revisited once openstack & qemu defaults change to virtio-scsi (if ever).
Test cases:
Qemu with `-drive if=virtio,format=$format,file=$image` should work, requires virtio-blk today.
Qemu with `-device virtio-scsi-pci,id=scsi0 \
-drive if=none,id=rootfs,format=$format,file=$image -device scsi-hd,drive=rootfs,bus=scsi0.0"` already works, requires virtio-scsi, and should continue to work.
Openstack with
openstack image create \
--container-format bare \
--disk-format qcow2 \
--file my-minimal.qcow2
should work and currently requires virtio-blk.
Openstack with
openstack image create \
--container-format bare \
--disk-format qcow2 \
--property hw_disk_bus=scsi \
--property hw_scsi_model=virtio-scsi \
--file my-minimal.qcow2
already works, and should continue to work, requires virtio-scsi. |
|
