Join GitHub today
GitHub is home to over 50 million developers working together to host and review code, manage projects, and build software together.Sign up
rm: skipping stage1 GC: device or resource busy due to 'upperdir is in-use by another mount' #3805
What did you do?
What did you expect to see?
What did you see instead?
Note, this doesn't actually prevent
There's a similar error which docker hits unreliably over here: coreos/bugs#2127
For reference, the full error is:
It comes from here:
Right, it doesn't. That's because stage0 logic tries to recover from complex issues with an "eventually clean" logic. This error however means that we are skipping the stage1 GC logic, which may result in other resources (i.e. non mounts/FS) leakage. I guess mostly cgroups and CNI items could be left around.
Partially true. It looks like the kernel commit you linked changed the behavior of mount() and broke this invariant in our code:
It used to be possible to double-mount the same overlayfs in place, but now that fails with EBUSY. We should proceed in that case, as it means we already have everything we need in place.