Le 9 février 2021 17:50:23 GMT+01:00, Lee Yarwood <lyarwood@redhat.com> a écrit :

On 09-02-21 12:23:58, Sean Mooney wrote:
On Tue, 2021-02-09 at 11:21 +0000, Arnaud Morin wrote:
Hey all,

From time to time we are facing an issue when puting instance in rescue
with the same image as the one the instance was booted.

E.G.
I booted an instance using Debian 10, disk are:

debian@testarnaud:~$ lsblk
NAME   MAJ:MIN RM  SIZE RO TYPE MOUNTPOINT
sr0     11:0    1  486K  0 rom  
vda    254:0    0   10G  0 disk 
└─vda1 254:1    0   10G  0 part /
debian@testarnaud:~$ cat /etc/fstab 
# /etc/fstab: static file system information.
UUID=5605171d-d590-46d5-85e2-60096b533a18	/	ext4
errors=remount-ro	0	1



I rescued the instance:
$ openstack server rescue --image bc73a901-6366-4a69-8ddc-00479b4d647f testarnaud


Then, back in the instance:

debian@testarnaud:~$ lsblk
NAME   MAJ:MIN RM  SIZE RO TYPE MOUNTPOINT
sr0     11:0    1  486K  0 rom  
vda    254:0    0    2G  0 disk 
└─vda1 254:1    0    2G  0 part 
vdb    254:16   0   10G  0 disk 
└─vdb1 254:17   0   10G  0 part /



Instance booted on /dev/vdb1 instead of /dev/vda1

Is there anything we can configure on nova side to avoid this
situation?

 in ussuri lee yarwood added
 https://specs.openstack.org/openstack/nova-specs/specs/ussuri/implemented/virt-rescue-stable-disk-devices.html
 to nova which i belive will resolve the issue
 
 from ocata+ i think you can add hw_rescue_bus=usb to get a similar
 effect but from ussuri we cahgne the layout so thtat the rescue disk
 is always used. lee is that right?

Yeah that's correct, from Ussuri (21.0.0) with the libvirt driver the
instance should continue to show all disks connected as normal with the
rescue disk appended last and always used as the boot device.

In your case, switching the rescue bus should change the default boot
device type n-cpu tells libvirt to use working around your problem.
Another obvious thing you can try is just using a different image so the
two MBRs don't conflict?

Cheers,