Storing AppVMs on Secondary Drives

Suppose you have a fast but small primary SSD and a large but slow secondary HDD. You want to store a subset of your AppVMs on the HDD.

R4.0

Qubes 4.0 is more flexible than earlier versions about placing different VMs on different disks. For example, you can keep templates on one disk and AppVMs on another, without messy symlinks.

These steps assume you have already created a separate volume group and thin pool (not thin volume) for your HDD. See also this example if you would like to create an encrypted LVM pool (but note you can use a single logical volume if preferred, and to use the -T option on lvcreate to specify it is thin). You can find the commands for this example applied to Qubes at the bottom of this R4.0 section.

First, collect some information in a dom0 terminal:

sudo pvs
sudo lvs

Take note of the VG and thin pool names for your HDD, then register it with Qubes:

# <pool_name> is a freely chosen pool name
# <vg_name> is LVM volume group name
# <thin_pool_name> is LVM thin pool name
qvm-pool --add <pool_name> lvm_thin -o volume_group=<vg_name>,thin_pool=<thin_pool_name>,revisions_to_keep=2

Now, you can create qubes in that pool:

qvm-create -P <pool_name> --label red <vmname>

It isn’t possible to directly migrate an existing qube to the new pool, but you can clone it there, then remove the old one:

qvm-clone -P <pool_name> <sourceVMname> <cloneVMname>
qvm-remove <sourceVMname>

If that was a template, or other qube referenced elsewhere (NetVM or such), you will need to adjust those references manually after moving. For example:

qvm-prefs <appvmname_based_on_old_template> template <new_template_name>

In theory, you can still use file-based disk images (“file” pool driver), but it lacks some features such as you won’t be able to do backups without shutting down the qube.

Example HDD setup

Assuming the secondary hard disk is at /dev/sdb (it will be completely erased), you can set it up for encryption by doing in a dom0 terminal (use the same passphrase as the main Qubes disk to avoid a second password prompt at boot):

sudo cryptsetup luksFormat --hash=sha512 --key-size=512 --cipher=aes-xts-plain64 --verify-passphrase /dev/sdb
sudo blkid /dev/sdb

Note the device’s UUID (in this example “b209…”), we will use it as its luks name for auto-mounting at boot, by doing:

sudo nano /etc/crypttab

And adding this line (change both “b209…” for your device’s UUID from blkid) to crypttab:

luks-b20975aa-8318-433d-8508-6c23982c6cde UUID=b20975aa-8318-433d-8508-6c23982c6cde none

Reboot the computer so the new luks device appears at /dev/mapper/luks-b209… and we can then create its pool, by doing this on a dom0 terminal (substitute the b209… UUIDs with yours):

First create the physical volume

sudo pvcreate /dev/mapper/luks-b20975aa-8318-433d-8508-6c23982c6cde 

Then create the LVM volume group, we will use for example “qubes” as the :

sudo vgcreate qubes /dev/mapper/luks-b20975aa-8318-433d-8508-6c23982c6cde 

And then use “poolhd0” as the (LVM thin pool name):

sudo lvcreate -T -n poolhd0 -l +100%FREE qubes

Finally we will tell Qubes to add a new pool on the just created thin pool

qvm-pool --add poolhd0_qubes lvm_thin -o volume_group=qubes,thin_pool=poolhd0,revisions_to_keep=2

By default VMs will be created on the main Qubes disk (i.e. a small SSD), to create them on this secondary HDD do the following on a dom0 terminal:

qvm-create -P poolhd0_qubes --label red unstrusted-hdd

R3.2

In dom0:

mv /var/lib/qubes/appvms/<my-new-appvm> /path/to/secondary/drive/<my-new-appvm>
ln -s /path/to/secondary/drive/<my-new-appvm> /var/lib/qubes/appvms/

Now, my-new-appvm will behave as if it were still stored on the primary SSD (except that it will probably be slower, since it’s actually stored on the secondary HDD).

  • The above procedure does not interfere with Qubes Backup. However, attempting to symlink a private.img file (rather than the whole AppVM directory) is known to prevent the private.img file from being backed up. The same problem may occur if the above procedure is attempted on a TemplateVM. [1]

  • After implementing the above procedure, starting my-new-appvm will cause dom0 notifications to occur stating that loop devices have been attached to dom0. This is normal. (No untrusted devices are actually being mounted to dom0.) Do not attempt to detach these disks. (They will automatically be detached when you shut down the AppVM.) [2]