CEPH d_type support?

[Zephrant]

I've created some VMs for clients that are on CEPH on a new test setup, and got back this from them:
Here's the error message docker gives when I explicitly set docker to use the overlay2 driver: failed to start daemon: error initializing graphdriver: overlay2: the backing xfs filesystem is formatted without d_type support, which leads to incorrect behavior. Reformat the filesystem with ftype=1 to enable d_type support. Backing filesystems without d_type support are not supported.

They are requiring that I somehow reformat the FS to enable d_type support for better performance in Docker. I'm not seeing any options in the create OSD command I used:
pveceph osd create /dev/nvme1n1

Is it possible to set fstype=1 when creating an OSD? I see from searches that sometimes it is =0 and sometimes it is =1. What is the driver for what that setting is created as?

 

[aaron]

I think there is a bit of a misunderstanding here. The VM that you run docker in, is stored on Ceph right? Ceph has nothing to do with how the VM formats its disk. It provides a block level semantic (RBD - rados block device) on top of the object storage that is Ceph. How the block level image (disk) is partitioned and formatted is the job of the VM itself. Now it might be hard to configure such file system parameters during installation (depending on the installer), but just to test it out, you can add an additional disk to the VM and then partition and format it by hand. This way you should be able to set the FS parameters as you need them to verify that it indeed helps.

 

[Zephrant]

Thank you, I was indeed confused. I didn't realize the installer for Centos 7 now defaults to XFS so the VM was running XFS on root, and that was the XFS that Docker was complaining about. I created a new VM on EXT4, and Docker is now happy.

 Storage Driver: overlay2
  Backing Filesystem: extfs
  Supports d_type: true
  Native Overlay Diff: true
  userxattr: false

Thank you for the help-