(Suddenly) cannot run xterm in a minimal disposable VM

User Support General
1

Hi,

I am experiencing a strange issue:

I have a (fedora-37) minimal named disposable qube (sys-dns, created as described in this post. I have always been able to run in dom0:

qvm-run -u root sys-dns xterm

Suddenly, this doesn’t work. The command above displays the standard Running 'xterm' on sys-dns and nothing happens - no XTerm window opens. Tried option -u user - same unfortunate result.

For sys-wall (the other minimal named disposable described in that same other post) and for the DVM template of sys-dns xterm works fine. Not for sys-dns though. FWIW, the DNS service works fine, i.e. the qube does its job. I just can’t open a terminal into it, so I am practically blocked from interacting with it. I suppose something (e.g. an update) may have messed things up as I have not touched anything on those qubes since I created them (at the time I wrote in the other thread).

How can I diagnose and fix this, please?

2

You can open the VM console from dom0 terminal with qvm-console-dispvm and check the logs.

3

I got this today on a normal StandaloneVM. I was able to make it work after restarting the qubes-gui-agent service inside the qube. Open a few terminal in that qube and then execute the following command in dom0:

qvm-run -u root <qubename> systemctl restart qubes-gui-agent

If everything opens instantly, then try to add more ram to that qube, that’s how I was able to fix it properly.

4

Thanks for the tips guys.

If everything opens instantly, then try to add more ram to that qube, that’s how I was able to fix it properly.

But I see it has enough free RAM:

[user@sys-dns ~]$ free
               total        used        free      shared  buff/cache   available
Mem:          957128      359516      356956        2172      240656      457348
Swap:        1048572           0     1048572
5

For some reason, it seems that the memory was filled on boot and it oom killed a few process on the way even with 400/4000 set for memory.

To get more information about your issue, check the logs coming from the qube here (dom0): /var/log/xen/console/guest-<qubename>.log

6

For some reason, it seems that the memory was filled on boot and it oom killed a few process on the way even with 400/4000 set for memory.

How do we know that?

FWIW, my sys-dns has 768/1024 memory setting and I don’t run hundreds of thousands of DNS requests per second, so I wonder why you suspect memory exhaustion.

To get more information about your issue, check the logs coming from the qube here (dom0): /var/log/xen/console/guest-<qubename>.log

I don’t see anything memory related in the log.

Today, when I run the terminal successfully, the log shows lines like:

audit: type=1130 audit(...): pid=1 uid=0 auid=... ses=... msg='unit=systemd-hostnamed comm="systemd" exe="/usr/lib/systemd/systemd" hostname=? addr=? terminal=? res=success'

Searching for similar lines (containing terminal=) with yesterday’s timestamp I still see res=success and there are no error messages around those lines.

Am I reading it wrongly?
What does your log show?

7

I was talking about my issue here because I found some “oom” line in the boot log. Since you’re experiencing the same behavior I thought it was the same problem. The increase in ram made my qube work again.

I can’t get them unfortunately since I automatically delete them everyday.