Description of problem (please be detailed as possible and provide log
snippests):

To support RBD mirroring between OCP clusters with overlapping networks (see https://issues.redhat.com/browse/RHSTOR-3866), the OSDs need to support binding to a service IP instead of the pod IP.

The mons already support this behavior with two command line parameters:
  --public-addr=<serviceIP>
  --public-bind-addr=<podIP>

The --public-bind-addr parameter was added for the mons to support Rook scenarios way back in 2017 with this PR: https://github.com/ceph/ceph/pull/16189

In the prototyping for RBD mirroring, the mons are responding on the global IPs both for traffic internal to the cluster and from external to the cluster. For more details, see the prototyping doc:
https://docs.google.com/document/d/1C5y8bX77B7YQUX4XSCKA2Gncg_BE_DViWVN83BUtPe4/edit#heading=h.dv90pv56ycn

For the global IPs to work with OSDs, now Rook needs to create a service with a global IP in front of the OSD, while the OSD daemon needs to still bind to the pod IP. Therefore, similarly to the mons, the OSDs also need to implement the flag:
 --public-bind-addr=<podIP> 


Version of all relevant components (if applicable):

This feature is in the 4.13 planning with: 
https://issues.redhat.com/browse/RHSTOR-3866


Does this issue impact your ability to continue to work with the product
(please explain in detail what is the user impact)?

This is necessary for the RBD mirroring scenario across OCP clusters with overlapping networks.


Is there any workaround available to the best of your knowledge?

No


Rate from 1 - 5 the complexity of the scenario you performed that caused this
bug (1 - very simple, 5 - very complex)?

3

Can this issue reproducible?

Yes, RBD mirroring doesn't work without this feature when there are overlapping networks.