- Retest
- Minor
TLDR: If you find that you still have leftover zfs datasets from when you had the ZFS storage Driver, run "zfs unmount dataset" and "umount /mount/location", then try removing the old docker datasets with "zfs destroy -r docker/dataset"
Actions taken:
Disabled docker in anticipation to switch from zfs storage driver to overlay2.
Update to v7 from last stable version.
Go to docker settings, used the "delete folder" and changed to overlay2
Compose up docker containers, everything looks fine.
Look around and noticed with "zfs list" that all the datasets created by the docker zfs storage driver still exist.
Stop docker. Delete docker folder contents. zfs destroy -r cpool/docker
nothing happens cli stuck processing
No logs other than confirmation of zfs destory command running
cancel zfs destroy command.
Try to stop array, nothing happens
Reboot
Upon boot, docker is still disabled. So I try to zfs destroy -r cpool/docker
still same issue, nothing happens
Try to delete just one docker dataset: zfs destroy -r cpool/docker/6a4f173927477f081f7cd22a560d2e3dd6bb82166190819ef083c0b309b73108
That works
check for open files: lsof | grep /mnt/cpool/docker
Nothing
make sure dataset is unmounted:
zfs unmount cpool/docker
umount /mnt/cpool/docker
umount /mnt/user/docker
zfs destroy -r cpool/docker
works
Configure docker again with overlay2 driver
OK
check zfs dataset
OK
The only thing I could come up with that I may have done... not let the destroy task run long enough. I did wait several minutes and this is on NVMe. So I don't think that's it, but still a small possibility.
But that still doesn't change that the "delete folder" action in the docker settings didn't delete the datasets and simply confirmed that the delete was successful. Maybe because I changed it to overlay2? So it didn't think it needed to run a zfs destroy command from the gui?
If you need more info, let me know!