In reply to Laurence's message of 2 Feb 2026:

CVMFS_REPOSITORIES="atlas,atlas-condb,grid,cernvm-prod,sft,alice"

Maybe you are missing cernvm-prod

The nut to crack was the missing mount of cvmfs-config.cern.ch to CVMFS.
Unfortunately no one has point out that this is necessary now.
Because of my system I have to register all the mounts manually to the automounter.

For now I have done my first Theory-docker wu.
Thank you Laurence

Docker container runs fine over here, but at the end all tasks fail:

Wed 04 Feb 2026 11:23:34 AM CET | LHC@home | [error] Can't rename output file slots/3/shared/output.tgz to projects/lhcathome.cern.ch_lhcathome/Theory_2922-4838179-486_0_r1805586945_result: rename() failed
Wed 04 Feb 2026 11:23:34 AM CET | LHC@home | Computation for task Theory_2922-4838179-486_0 finished
Wed 04 Feb 2026 11:23:34 AM CET | LHC@home | Output file Theory_2922-4838179-486_0_r1805586945_result for task Theory_2922-4838179-486_0 absent

Is it an issue on my side?

Did likely have the same problem with Docker on Arch linux(final steps fail and on the LHC website no Stderr output)

https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=6438&postid=52901

Not sure why docker did not work but Podman with linger enabled works fine with multiple tasks finished successfully.

Theory-docker tasks run fine now.
But run-time and cpu-time are far from realistic!

Podman runs ok,
On my side not with lingering but containers switched to cgroupfs

cgroup_manager = "cgroupfs"

In reply to Saturn911's message of 4 Feb 2026:

Theory-docker tasks run fine now.
But run-time and cpu-time are far from realistic!

The run-time is realistic. Don't look at BOINC Manager's "Remaining (estimated)" time. That was and will always be way off, because Theory tasks have a wide variaty of runtimes.
Cpu-time: For a lot of systems that time is multiplied with the number of host cores. That bug is fixed in the next version of docker_wrapper.

I posted to the BOINC forum if its even possibale to use systemd. I have not found a way to make it work.

I don't need linger with cgroupfs, systemd & linger does not work for me.

In reply to Toby Broom's message of 4 Feb 2026:

systemd & linger does not work for me.

+1

While using systemd with Gentoo, I'd always get this error:

error running container: from /usr/bin/crun creating container for [/bin/sh -c dnf install -y https://cvmrepo.s3.cern.ch/cvmrepo/yum/cvmfs-release-latest.noarch.rpm && dnf install -y epel-release && dnf install -y libxcrypt-compat bc bzip2 lighttpd procps-ng make gcc which cvmfs bind-utils netcat zlib-devel compat-openssl11 freetype && dnf clean all]: sd-bus call: Access denied as the requested operation requires interactive authentication. However, interactive authentication has not been enabled by the calling program.: Permission denied
: exit status 1

I've found that this formula works with systemd and rootless podman. If you setup a user level podman service and podman socket in the regular account you want to use for boinc and then run your boinc client thru that user account, the podman app works with boinc.

systemctl --user enable podman.socket podman.service

systemctl --user start podman.socket podman.service

Then start boincmgr in your desired directory in the regular user account you're using. Because it's a regular user and not the "boinc" service user, it has a proper $XDG directory structure set up to run the containers using the user privileges from your regular user account. It even stopped complaining about falling back to --cgroup-manager=cgroupfs. Just remember to give your regular user account subuids and subgids.

I will note that I don't have to set up the boinc client this way with BOINC-Central@home. It can run podman thru the BOINC service daemon.