Correction: Chimaera is affected, not Beowulf. In both my Beowulf installs anacron isn't installed. It is on both Chimaera installations.

In crontab I just find the usual @reboot script that connects my nfs shares on my file server.

The effect is sporadic.

rolfie

There is always a slight delay during "Waiting for /dev to become fully populated", that normally takes a few seconds.

rolfie

I have no idea how to access the packages on a Devuan DVD correctly. In the internet there should be instructions how to address such a DVD in the sources.list. I consider this to be the pre-requisite for a successful installation of any package from that DVD. I guess thats missing in your case. Unfortunately I never had to go this path, I can't help.

In addition I searched Debian.org for the a package called linux-source. I never came across that. This really exists, its a meta-package that adresses the latest kernel sources, not just the headers. There also is a linux-source-4.19 meta-package. When the DVD is set up correctly in the sources.list, you should get the matching source for your kernel.

rolfie

Well, I have used a different way to install VBox on my stock Beowulf installations (with backports kernels). Loaded the Debian10 package down to /tmp, installed it with dpkg -i followed by a apt -f install to correct all dependencies. No trace of any dbus errors. My latest learning: maybe an apt install ./package.deb takes care for the dependencies too.

In addition, I added the vboxdrv, vboxnetadp and vboxnetflt to the /etc/modules file. 

I have not seen any need to fool around with update-rc.d. Maybe the modules did the job.

Besides the installation of the extension pack, the USB passthrough depends on that the user using the machine is member of the vboxusers group. Make sure you/the user are member of this group. The group is created automagically, but the user(s) needs to be added manually.

Full ACK from my side.

rolfie

Try apt -f install ....

rolfie

You forgot the most important thing: "*"

Have a very close look at what I did post and compare to yours.

rolfie

I have:

# apt list lsb*
Auflistung... Fertig
lsb-base/stable,stable,now 10.2019051400 all  [installiert]
lsb-release/stable,stable,now 10.2019051400 all  [Installiert,automatisch]

rolfie

Got the 3.1.0 with openrc to work, Cinnamon as desktop, efi mode, /boot unencrypted, encrypted LVM for root and swap, separate home.

Despite having a mirror set up, the CDROM still was present in the sources.list. Easy to fix.

The installation booted grub with shimx64.efi. I removed it afterwards.

Locale was fine with this setup.

Would say that runit is experimental in the current status.

rolfie

Update on the Beowulf 3.0.0 install on P4/P5:Same issues basically, the boot also ends in the grub-console, no grub menu shown.

Started the rescue mode and fixed issue #1. Looked at the locale, also is at C.UTF-8. Then fidelled a bit with boot, grub-install and update-grub, grub.cfg, efivars and the efibootmgr. Finally used efibootmgr -v and saw that all Devuan entries in the efi memeory were referring to shimx64.efi, despite my disabling Secure Boot.

Well, searched the man pages for grub-install and found the option --no-uefi-secure-boot.
A grub-install --no-uefi-secure-boot --bootloader-id=devuan4 brought a grubx64.efi into play. A reboot showed the usual blue grub splash screen, I am now working on P4/P5 and the Chimaera upgrade. The locale issue is gone too.

Next step is trying the same on the 3.1.0 install.

What I am wondering about: how did I get into this situation in the first place?

rolfie

It may not be that simple. All depends on if the meta package linux-image-amd64 is installed, and if this is linked to the main repository or to backports.

If you choose linux-image-amd64 during installation and not one of the presented dedicated kernel images, apt update && apt upgrade will check for updates and install them when available. For Beowulf this is the 4.19 kernels.

When you only install one dedicated kernel from backports, that will stay forever and not be affected by apt update && apt upgrade. Unless you install the meta-package from backports. Then the stock kernel is left alone and backports is updated.

You can't have both.

rolfie

Read this thread: https://dev1galaxy.org/viewtopic.php?id=3616

My suspicion: on a crash the kernel overwrites contents of the efi flash. Got an older socket AMD3+ board that had a similar issue (https://dev1galaxy.org/viewtopic.php?id=2820).

Now all my installations have /sys/firmware/efi/efivars/ set to ro in fstab.

I had the ASUS Prime X470-Pro sent back to service. They flashed a new bios and fixed the efi memory, something I could not achieve with the means available to a normal user. Now the board works normal again.

rolfie