I upgraded a 13.0.25 installation to .28. This installation has been working on Docker since it became mandatory to run it as a Docker image.
After the upgrade, the seafile daemon crashes at 23:55, which after some post mortem appears to be a spot where the logs are rotated. The crash leaves nothing in the logs as seafile.log and seahub.log are both 0 sized at that point.
Comparing to a working system, its seafile.log and seahub.log are owned by 8000:8000 which appears to be in line with plenty of other stuff too. In this one it was owner by root:root. Chowning the files to 8000:8000 makes the system start again.
As I had an earlier image of this virtual machine, I restored it elsewhere and examined it. That one had the files owned by root:root also earlier, but it was chmodded to be writable by other. Note that I have not touched the log files in any way, they got created as-is when the docker image was first deployed.
So my question is, was something changed in .28 to cause this?
I compared logrotate.logs on working and this installation, they have a difference when it comes to these specific files:
broken:
creating new /opt/seafile/logs/seafile.log mode = 0644 uid = 0 gid = 0
working:
creating new /opt/seafile/logs/seafile.log mode = 0644 uid = 8000 gid = 8000