Bug #16754
closedmounting cephfs root and sub-directory on the same node makes the sub-directory inaccessible from root.
0%
Description
ems@host1: sudo mount -t ceph 10.242.42.216:/ /mnt/mycephfs/
ems@host1: sudo mkdir sub1 & sub2
ems@host1: /mnt/mycephfs $ ls
sub1 sub2
ems@host1: sudo umount /mnt/mycephfs *************************************************************
ems@host1: sudo mount -t ceph 10.242.42.216:/sub1 /mnt/mycephfs/
ems@host1: sudo mount -t ceph 10.242.42.216:/ /mnt1/cephfs/ *************************************************************
ems@host1: cd /mnt1/cephfs/
ems@host1:/mnt1/cephfs$ ls
ls: cannot access sub1: No such file or directory
sub1 sub2
ems@host1:/mnt1/cephfs$ ls /mnt1/cephfs/
/ sub2
Similarly if we mount sub2 on a host2, and then also mount the root to host2, then Sub2 is not accessible from root mount point.
Updated by Zheng Yan almost 8 years ago
I can't reproduce this on 4.5 kernel. which verion of kernel are you using?
Updated by John Spray almost 8 years ago
- Status changed from New to Can't reproduce
Closing because this is happening on an old kernel. Please re-open if you can reproduce the issue on a recent 4.x kernel.
Updated by Rohith Radhakrishnan almost 8 years ago
@John: For Ubuntu 14.04.02, 3.16v kernel is the latest and I using the most recent version of 3.16 kernel. In that case shouldn't this kernel version be supported?
Updated by John Spray almost 8 years ago
@Rohith new Ceph patches go onto the latest upstream linux kernel, and then it's up to linux distributions which things they choose to backport to their stable kernels. I don't know if Ubuntu kernel maintainers backport anything for cephfs, you could ask them.
Updated by Greg Farnum almost 8 years ago
Hmm, we aren't very good about tagging CephFS patches for stable trees. Distributors might have an easier time maintaining kernels once we get better at that.
(Although I think we also have a lot of big "bug fix" patches that aren't really suitable for stable trees, but hopefully less of those as time goes on.)
Updated by Rohith Radhakrishnan almost 8 years ago
@Greg Farnum does that mean there is a chance that 3.16 kernel may never have the fix back-ported ?
Updated by Greg Farnum almost 8 years ago
Given the current state of CephFS kernel support I wouldn't expect so. The upstream community still asks users to run the newest kernel they possibly can and most distributions aren't doing their own backporting AFAIK.