OpenStack allows to attach multiple floating ips to a single server.
Previously, only one floating ip was supported by this module. It
would call openstacksdk's get_server_public_ip(), which in turn
would return just one of the attached floating ips. If this floating
ip would not point to the right nat_destination or fixed_address,
then the module would fail.
If no floating ip had been attached to a server, then this module
would call openstacksdk's add_ips_to_server() with both parameters
"floating_ip_address" and "network" to attach a floating ip to the
server. But both parameters are mutually exclusive [1], i.e.
add_ips_to_server() will ignore "floating_ip_address" if "network"
is set and then choose any non-attached floating ip from "network".
If "floating_ip_address" has not been created in OpenStack and
"network" is not given, then this module would not create this
floating ip [2].
The new module code allows to create and add more than one floating
ip to a server. It priorizes more specific parameters over generic
ones, i.e. if both "floating_ip_address" and "network" are given,
then "floating_ip_address" precedes "network".
Parameter "network" is now required if "floating_ip_address" is
specified, because both are necessary when creating floating ips.
Module documentation and args have been updated accordingly.
Ref.:
[1] a6b0ece282/openstack/cloud/_floating_ip.py (L987)
[2] a6b0ece282/openstack/cloud/_floating_ip.py (L907)
Task: 40939
Story: 2008181
Change-Id: I1ada1be0994f526f72f81f7458782afbcca3c92c
Previously, when updating a router all its interfaces where removed and
readded by Ansible's os_router module. As a unwanted side effect all
active connections of the router and nat'ed devices where dropped,
closing e.g. all active tcp sessions. Now, only necessary changes
are applied.
Task: 40136
Story: 2007845
Change-Id: I172caf360e6e342dd54865da5a5b72b0dc0205c8
In order to use a newly generated pool for a new
healthmonitor the pool must be online and active,
to avoid 409 conflict, when trying to update a pool
with e.g pending update as a status.
Change-Id: I160a75c6fbbf1555f3adcb444c77057f59b4cdfb
Previously, all subnet properties were updated if any value did not match.
But this behaviour caused errors like e.g. "Current gateway ip 192.168.0.1
already in use by port [ID]. Unable to update." even if that gateway was
not about to be changed.
Task: 40927
Story: 2008172
Change-Id: I049b0dade4c7ea3e1ef24777ae558f650caa136c
