ERRO[0000] User-selected graph driver "overlay" overwritten by graph driver "vfs" from database - delete libpod local files to resolve

[adrianreber]

CRIU is running Podman based checkpoint/restore tests in Travis. Because the Travis kernel is broken for checkpoint/restore on overlayfs we switched to VFS:

https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1857257

To tell Podman to use VFS I am running podman --storage-driver vfs info first before running any other Podman command. This used to be good enough to initialize Podman to use VFS. This still kind of works, but now, every time Podman is started I see following error message:

ERRO[0000] User-selected graph driver "overlay" overwritten by graph driver "vfs" from database - delete libpod local files to resolve

Everything works as expected but now we get a lot of those error messages in CI.

This is using 6-podman_1.7.0~6_amd64.deb from the kubic stable repository for 18.04. The complete log can be found here:

https://travis-ci.org/checkpoint-restore/criu/jobs/646854308

Right now this seems to be purely cosmetic, but it is new and used to be not visible.

Probably I am using the wrong approach to switch Podman to VFS, but that is how it used to work before.

[mheon]

This is the result of storage.conf still having overlay set, while the database contains vfs from your initial podman info command. The database wins, but this could be very confusing for users (your configuration file is being overridden), so we write errors.

[adrianreber]

Is the only way around this modifying storage.conf?

[mheon]

I think removing it would also work, but that might prevent us from locating the correct base directory for Podman files.

[vrothberg]

ERRO[0000] User-selected graph driver "overlay" overwritten by graph driver "vfs" from database - delete libpod local files to resolve

The error message seems to be swapping the database and user-selection as well.

Why are we logging an error for this case? Wouldn't an Info log be sufficient?

[vrothberg]

Cc: @giuseppe @rhatdan

[github-actions]

A friendly reminder that this issue had no activity for 30 days.

[rhatdan]

@adrianreber Any update on this issue?

[adrianreber]

I do not see the error message any more in the current CRIU CI runs based on Podman 1.8.1~1.

Closing it.

[JJ]

Same error in podman 2.2.1:

User-selected graph driver "overlay" overwritten by graph driver "vfs" from database - delete libpod local files to resolve

[rhatdan]

@JJ Please open new issue or follow the instructions above to cleanup the problem.

Bottom line is you ran podman with vfs mode, and now this is recorded in the database. Only way to switch is to reset your storage or delete the libpod database.

[PhilipSchmid]

For all wondering what @rhatdan probably meant with "reset your storage":

sudo rm -rf ~/.local/share/containers/

At least it worked for me on Pop_OS 20.04 with Podman (3.0.0) 😄 ...

[Kiritow]

If you are using nested podman, this might help: rm /var/lib/containers/storage/libpod/bolt_state.db