systemd-journald race condition due to /etc/machine-id regeneration

[jtetrault]

Hey, I think this is the appropriate repo for this issue (at least, it can be resolved in this repo).

There seems to be a minor race condition on first boot, that causes journalctl to just spit out "No journal files were found" until the service / system is rebooted. I traced this down to /etc/machine-id being deleted and regenerated by the generic-startup.sh script in the boot-scripts repo.

Basically, this is what happens:

• For whatever reason, /etc/machine-id is present in the final generated flasher image
• After flashing, on first boot, journald-systemd grabs the machine-id found at /etc/machine-id, and starts logging to /var/log/journal/<machine-id>
• generic-startup.sh runs, sees the /etc/ssh/ssh.regenerate file exists, and removes and regenerates the machine-id
• journald-systemd is now logging to the "wrong" spot for the new machine-id, so journalctl is unable to find the log files

I deleted /etc/machine-id from a flasher sd card, used it to flash a BBE, and did not see the above problem on the first boot. So I think all that's required is to add a rm -f /etc/machine-id somewhere in chroot.sh

[RobertCNelson]

Thanks @jtetrault i just added to the same location as teh ssh trigger file..

[RobertCNelson]

also fixed the flashing script to do the sam.e..

[jtetrault]

Thank you!