I've tested this out and it seems good. Tried deploying then scaling for
ceph, cinder, glance and nova-compute and all had updated ceph.conf on the
client side. The use of request-id combined with unit-name scoped responses
seems to nicely avoid any collisions as well. I have a few comments on some
minor fixups inline.
I've tested this out and it seems good. Tried deploying then scaling for
ceph, cinder, glance and nova-compute and all had updated ceph.conf on the
client side. The use of request-id combined with unit-name scoped responses
seems to nicely avoid any collisions as well. I have a few comments on some
minor fixups inline.