trove-mgmt-cli's "host > update_all" seems to claim to be able to upgrade/update the guest on deployed hosts, but it does not work as advertised.
I see https://github.com/openstack/python-troveclient/blob/b98893926d14bd0b5b2706c808688659695c1731/troveclient/hosts.py#L49-L54
however, I can't find a corresponding "update" action in the mgmt api:
https://github.com/openstack/trove/blob/166fd33ee013488bca13507fba03da73697f0199/trove/extensions/mgmt/instances/service.py#L97
There's no mention in the apischema of an update/upgrade action, so I'm thinking that this feature wasn't fully ported. On a semi-related note, trove-mgmt-cli's "host > list" exposes this fault: https://bugs.launchpad.net/trove/+bug/1130983. The gerrit review for the proposed fix (https://review.openstack.org/#/c/32916/) had this comment added before the commit was ultimately abandoned:
> ""rdhosts" is a Rackspace Nova API extension. IIRC Mike put those in the
> public at some point as a fork, although we need to
> probably find a better way to handle them.
>
> We *should* change things to work with standard Nova functionality, but
> this function has been changed so that its return values > are entirely
> different. This will break expectations of Rackspace ops (as well as tests)
> are relying on. I think a better solution is to > "fix" this in a way that
> doesn't mean radically changing the views, or simply creating a new API
> call entirely."
Id rather fix the existing api calls if possible.