LHC@home

VirtualBox 7.0.2 (released October 20 2022)
https://www.virtualbox.org/wiki/Changelog-7.0#v02

This is a maintenance release. The following items were fixed and/or added:

Known issue: VMs having more than one vCPU configured will not work properly on macOS Catalina due to an unknown memory corruption issue. Either lower the number of vCPUs to 1 or upgrade to BigSur or later where the issue does not occur
Main: Fixed issue when VBoxSVC could become unresponsive if Extension Pack was not installed (bug #21167)
macOS hosts: Added workaround for a bug in the Hypervisor framework on Catalina causing VERR_NEM_MAP_PAGES_FAILED errors when starting a VM. (bug #21128)
macOS hosts: Re-introduced support for internal networking, this is considered a bit experimental still
macOS hosts: Fixed VM crash when the guest tries to access a microphone or webcam
Windows host: Shared Clipboard: Fixed issue when only 4Kb of host clipboard buffer was accessible to guest (bug #21149)
Linux Guest Additions: Introduced initial support for kernel 6.1
Linux Guest Additions: Fixed issue when VBoxClient seamless service caused a crash of some X11 applications (bug #21132)
Windows hosts: GUI: Fixed missing Qt libraries for vista style and sql driver (bug #21155)
GUI: Fixed a glitch in the log viewer which was causing wrong log file to be saved (bug #21156)

For me its not super reliable I get this error.

Do you have a link to the task causing it?

On the dev-system I had the same VBoxHeadless-error (twice) with VBox 7.0.0:

https://lhcathomedev.cern.ch/lhcathome-dev/forum_thread.php?id=597&postid=7827

This kind of error appears in the virtualbox forum every now and then for years independent of the virtualbox version.
It looks like it has to do with "restoring a saved state".


Some suggestions based on the posts on the virtualbox forum.

1.
Before any vbox upgrade not yet started tasks should be set on "hold" and running/paused task should run until they are completely finished.
Then remove the old vbox version, reboot, install the new version and reboot again.

2.
In some cases 'VirtualBox.xml' must also be removed (manually) before the new version gets installed.
Be aware:
Each user account has it's own 'VirtualBox.xml' which may contain a list of user created VMs.
Those VMS also must not be in "saved state".
Instead they must completely be shut down.
Removing 'VirtualBox.xml' does not delete those VMs but clears the list in vboxmanager.
Hence, the VMs have to be attached again to the new vbox version.

3.
It looks like in some cases parts of the old vbox version (files, reg-keys) remain on the computer which always leads to trouble sooner or later.
They need to be removed before the first reboot from (1.),

It's a bit hard to pin down if I have 46 tasks running at the same time. Any tips?

1. I aborted all the task and it came back, maybe as you say after reboot, so they would be restored from saved state.

I didn't have any problems for a long time with 6.1, some issues I imagine with the 7.0 branch as it's new.

I go back to 6.1 branch

Oracle VirtualBox 7.0+ seems not 100% reliable. Another crash of VBoxHeadless.exe during (re)start of a VM.

https://lhcathome.cern.ch/lhcathome/result.php?resultid=368887353

What makes me wonder are these logfile lines pointing out a BOINC client crash during the first run of that task:

10:31:20 (5956): BOINC client no longer exists - exiting
10:31:20 (5956): timer handler: client dead, exiting

'client dead' means the BOINC client.
The 2nd run may have crashed due to trash left behind by the 1st run.

Yeah, the BOINC client crashed suddenly after suspending 2 WCG-tasks, but there was no trash left behind for the VM.
The VM powered off successfully. Thereafter I rebooted that system.
I was running that Theory with snapshots and the VM was restored OK from the snapshot.
When starting the restored VM (by the wrapper of course), VBoxHeadless.exe crashed like I mentioned twice when testing on LHCs dev-system.
Like back then a Windows popup with VBoxHeadless.exe crash appeared.