Post Fedora-32 Template Installation

[Qubes]

I installed a clean Fedora-32 template today, but I have some issues and
I am not sure where to start.

After installing the F-32 template I switched the default-mgmt-dvm qube
to use the F-32 template.

I then created a new qube basing it on the F-32 template,
test-fedora-32-template, just to check that everything is working ok on
this test vm before changing all of my F-31 based qubes over to F-32.

I am unable to boot the test-fedora-32-template qube. After a while I
receive a system notification, the ones that pop up top right of the
screen, saying,

'Cannot connect to qrexec agent for 60 seconds, see
/var/log/xen/console/guest-test-fedora-32-template.log for details'

and if I look in the log file I see hundreds of messages like this...

'.....Failed to fork() worker: Cannot allocate memory'
'.....systemd-udevd[272]: xvdc1: Failed to process device, ignoring: No
space left on device'
'.....eth0: Failed to process device, ignoring: No space left on device'

The F-32 template itself starts up without any problems.

[Qubes]

The problem described above definitely has something to do with the f-32
specifically.

Even after changing my test qube to the f-31 template and starting it
the qube still refuses to boot and gives me the same issues in
/var/log/xen/console/guest-test-fedora-32-template.log.

However, if I create a new qube based on my f-31 template then
everything works as expected.

The after effect of this is also, because I changed my default-mgmt-dvm
to the f-32 template, in order to update the f-32 template, which
subsequently failed because machines based on the f-32 template don't
work, my default-mgmt-dvm is also bust. I guess I can recreate it, but
that still does not solve my f-32 problems.

AND, all of these issues started when I accidentally interrupted the
f-32-minimal template installation as I described in a previous email to
this group 'Template Installation Problem'.

In the meantime I have managed to get the minimal template installed
using the --action=reinstall option as described in the documentation.

'sudo qubes-dom0-update --action=reinstall qubes-template-fedora-32-minimal'

Following on from my previous email I also had to reinstall the f-32
template because it gave the same issues I I describe here now. I hoped
that reinstalling the template would fix the issues, but they appear
deeper than the template.

I really hoped I would be able to fix this without having to resort to
using a full system backup and starting from 'scratch'.

[Qubes]

I finally got this one figured out as well. It's been a good week! :-)

So it turns out that when you set the memory parameters for a Fedora-32
template you cannot set the initial memory lower than 400 MB. If you set
the initial memory parameter to 300 MB for example the template, and any
AppVM's based on the template, will fail to start. The errors that are
shown to the user and the event logs don't indicate that the problem is
with initial memory being to low.

Perhaps the devs can put a error check that will prevent you from
settings the initial memory below 400 MB, exactly like we currently have
a check that prevents you from setting initial memory below 10% of maxmem.

In hindsight it may be a silly error but it never crossed my mind that
300 MB initial memory can cause a VM not to boot. Fedora-31 templates
boot with initial memory set to 200MB so I simply continued with the
same parameter when I started with Fedora-32.