Closed apinter closed 2 years ago
Thanks for the report.
From a quick skim, this may have the same root-cause as https://github.com/coreos/fedora-coreos-tracker/issues/1138.
The fix for that went into testing
stream release 36.20220522.2.1
.
@apinter It would be great if you could quickly try on a testing
node and check whether that fixes your issue.
@lucab You're correct, switching to testing did solve the problem. Going to keep the system rolled back to the previous stable release with F35 and Podman 3 till this fix makes it to the stable stream. Big thanks for the help!
Describe the bug The upgrade to F36 is bringing Podman 4 which for some reason broke my containers. I do have an additional disk formated to
btrfs
that functions as storage for the containers (grep graphroot /etc/containers/storage.conf graphroot = "/var/lib/gitlab_data"
&&grep driver /etc/containers/storage.conf driver = "btrfs"
). Existing containers can't start, new containers are the same.Reproduction steps Steps to reproduce the behavior:
systemd
or launch a new one for example:podman run --rm -it --log-level debug fedora:36
systemd
:Expected behavior Containers start without issues, business as usual.
Actual behavior Containers fail to start with a
read-only file system
error which the fs is not. Everybtrfs
subvol podman creates isrw
.System details
Fedora CoreOS version:
Ignition config Empty fields means redacted information due to sensitivity.
The above systemd units may not match reality, since then I did generate units with podman (this is about a year old server), for example:
Additional information
What I could do - which would be a pain - is to migrate the entire disk to a new disk that is "compatible" with the new Podman version if required. This works though, but not a desired solution. Would be happier if I would know what went wrong and how could it be mitigated.