It has been proposed that either using a mount file unit file, such as: 
-----------------------------
[Unit]
Description = Mount file system that lives on /VDO
name = VDO.mount
Requires = vdo.service 
Conflicts = umount.target
 
[Mount]
What = /dev/mapper/vdo
Where = /VDO
Type = xfs
Options = discard
 
[Install]
WantedBy = multi-user.target
-----------------------------

instead of the fstab entry will correct this issue

or, modifying the vdo.service file to:

-----------------------------
[Unit]
Description=VDO volume services
After=systemd-remount-fs.service
DefaultDependencies=no
Conflicts=shutdown.target
Before=local-fs-pre.target

[Service]
Type=oneshot
RemainAfterExit=yes
ExecStart=/usr/bin/vdo start --all --confFile /etc/vdoconf.yml
ExecStop=/usr/bin/vdo stop --all --confFile /etc/vdoconf.yml

[Install]
WantedBy=local-fs-pre.target
-----------------------------

With this change to the vdo.service file there is no longer a need for 
adding x-systemd.requires=vdo.service in the fstab file.

It is important that VDO doesn't delay or prevent the system from starting up
if there is a rebuild of the volume.  If the process of VDO rebuilding prevents other services from starting using a different "Before" target then the solution might not be favorable.   

Here is the current vdo.service configuration

# /usr/lib/systemd/system/vdo.service
[Unit]
Description=VDO volume services
After=systemd-remount-fs.service

[Service]
Type=oneshot
RemainAfterExit=yes
ExecStart=/usr/bin/vdo start --all --confFile /etc/vdoconf.yml
ExecStop=/usr/bin/vdo stop --all --confFile /etc/vdoconf.yml

[Install]
WantedBy=multi-user.target

-----------------------

Some testing will be required.

Note: the vdo.service unit file has not been tested yet.

Adding this change to vdo.service file makes mounting VDO volumes easier and
fixes the problem of "Breaking order cycle", but adds a "cost" of delaying the
boot cycle if the VDO volume has not been cleanly shutdown.

VDO volumes perform a recovery after unclean shutdowns.  If the vdo.service file
is changed to use "local-fs-pre.target", then the system will wait until VDO has
completed it's recovery before proceeding.   The time it takes to finish a
recovery can depend on the VDO volume size, speed of the CPU and media.

Simulating this could be done by creating a VDO volume on a USB thumb drive
(slow media), writing data to that device, then crashing the system.  After
the system has been rebooted, run systemd-analyze blame to see how long VDO
took to start the volume(s).  In my test it took about 90 seconds. During the recovery process the system can't be accessed, since systemd hasn't reached the
point of starting SSH or other services to manage the system.

Using the mount unit file instead of changing the vdo.service fill will mount
vdo volumes after the system has booted and doesn't cause the "Breaking order cycle".

Pro/Con
1. vdo.service
   - would make mounting VDO volumes easier for many users
   - But VDO recoveries can extend the boot cycle
2. mount unit
   - adds additional steps, which are not intuitive to everyone
   - VDO does not get in the way of the boot up process or system access

Adding the proposed vdo.service file can delay the boot cycle when VDO volumes
need to recover, therefore we should not implement this feature.  
Users should create mount unit files for VDO volumes unless there is an alternate 
method of initializing vdo volume that doesn't delay the boot or break ordering.

What about adding both x-systemd.requires=vdo.service and _netdev to /etc/fstab?

That should also get rid of the cycle and doesn't require people (and Cockpit) to switch from /etc/fstab to something new.

_netdev may allow for this to work, but we're not really a network device, so it doesn't fit the definition exactly.  I think it would be incorrect for that reason.

Maybe we can add a documentation note in the  Storage Admin guide to specify that option noting that we're not a network device, but it achieves the goal for SELinux compliance?

If someone had iscsi storage and created a VDO volume on top, then a filesystem, then NFS, would using _netdev have problems figuring out which network device / service to start first (nfs, vdo, iscsi)?

This issue is resolved by creating a mount unit file

 https://bugzilla.redhat.com/show_bug.cgi?id=1552242

While using the "Before" option mentioned above can also resolve this issue it can
have the side effect of extending the boot process.

Closing ticket