Ceph rbd volume stuck in use

[kfox1111]

We hit bug:
ceph/ceph#50861

A watch wouldn't release and so a volume was stuck unmountable.

The fix was to run rbd info on the volume, then find its block_name_prefix property, then use that in:
ceph osd map <block_name_prefix>
which gets something like:
osdmap e352096 pool 'xxx' (44) object 'rbd_header.abee677ba13b5d' -> pg 44.a297414 (44.14) -> up ([57,91,163], p57) acting ([57,91,163], p57)

And then you can restart the primary osd (57 in the above example)