Qubes-rootfs-resize.service prevents boot

skyvine · March 25, 2025, 9:12am

Whenever I try to start a qube that I recently resized, it fails with a message that it could not connect to qrexec within 60 seconds. When I open the log at /var/log/xen/console/guest-VMNAME.log I see messages like this:

Job qubes-rootfs-resize.service/start running (3s / no limit)
Job qubes-rootfs-resize.service/start running (4s / no limit)
...
Job qubes-rootfs-resize.service/start running (57s / no limit)

This messages appear on every boot and cut out close to the 60s mark… so I’m inferring that the resize service blocks the qrexec service causing dom0 to give up every time it boots.

Not gonna lie, I’m not sure how to get out of this one.

phceac · March 25, 2025, 9:49am

I just noted this in my “qubes tips”…

You can increase the timeout using a command in Dom0 :

 `qvm-prefs <vmname> qrexec_timeout 3600`

Only replace <vmname> to get a delay of one hour, if that is enough.

kenosen · March 25, 2025, 1:21pm

github.com/QubesOS/qubes-issues

VMs fail to start because adjusting filesystem size times out

opened 05:54PM - 22 Oct 18 UTC

SuzanneSoy

C: core P: major ux r4.0-jessie-stable r4.0-stretch-stable r4.0-centos7-stable r4.0-buster-stable r4.0-fc27-stable r4.0-fc28-stable r4.0-fc29-stable r4.1-buster-cur-test r4.1-stretch-cur-test r4.1-fc29-cur-test needs diagnosis regression affects-4.2

### Qubes OS version: 4.0 ### Affected component(s): VMs b…ased on Fedora-26, likely debian-9 too. --- ### Steps to reproduce the behavior: Some VMs fail to start: `Cannot connect to qrexec agent for 60 seconds, see /var/log/xen/console/guest-xxx.log for details`. In that log file, I can see that the problematic VMs are busy with the following task (the time updates until around 55s when the VM is killed): ``` A start job is running for Adjust r…t filesystem size (45s / no limit) ``` ### Expected behavior: The VMs should start, even if that takes 10 minutes. Beyond 60 seconds it is reasonable to show a message and invite the user to check the logs to see what's going on. ### Actual behavior: The VM gets killed, preventing access to its data and applications. ### General notes: After seeing this commit https://github.com/marmarek/qubes-core-agent-linux/commit/7b4d453a23b066e5919f5498c5b3c75f8df2bae1 I looked for scripts named `*resize*` in `/usr/lib`, found one (`resize-rootfs-if-needed.sh` I think, but there are a couple), and disabled it (added `exit 0` at the top). This allowed me to boot the VMs. I grepped through QubesOS's code but couldn't find that 60-second timeout is defined, 60 is a fairly common number it seems :) . --- ### Related issues: Related to the fix at https://github.com/marmarek/qubes-core-agent-linux/commit/7b4d453a23b066e5919f5498c5b3c75f8df2bae1 for #3173 and #3143.

skyvine · March 26, 2025, 7:31am

Thanks for the tip! It succeeded after 3 minutes.

phceac · March 26, 2025, 9:45am

Nice !

It might be wise to put it back to default. I think the command is :

 qvm-prefs VMNAME -D