On September 1st, 2015, 9:36 a.m. UTC, Thomas Woerner wrote:

src/rolekit/server/io/rolesettings.py (Diff revisions 6 - 7)
		63	elif name in self.get_instances(self._type):

This is resulting in a ValueError for already deployed roles (instances) while roled starts.
The ValueError should only happen if there is another instance or pendingrole with the same name. It should not complain on a conflict the instance iteself.

2015-09-01 11:08:39 DEBUG1: Loading 'foo' instance '1'
2015-09-01 11:08:39 Traceback (most recent call last):
  File "rolekit/server/decorators.py", line 40, in handle_exceptions
    return func(args, *kwargs)
  File "rolekit/server/dbusrole.py", line 80, in init
    settings = RoleSettings(self._name, instance)
  File "rolekit/server/io/rolesettings.py", line 64, in init
    raise ValueError("Instance name already in use")
ValueError: Instance name already in use

- Thomas

On August 31st, 2015, 3:37 p.m. UTC, Stephen Gallagher wrote:

We want to be able to support configuring role deployments during
automated system setup (such as Anaconda kickstart for Fedora and
Red Hat Enterprise Linux systems). With this patch, we add a
--next-boot argument to 'rolectl deploy'. This will cause a one-
time systemd service unit to be created and configured to run at
the first boot.

This service unit will execute the real 'rolectl deploy' after the
network is online and will process the role deployment, then
delete the temporary files. These files will not be deleted in the
event of an error.

Fixes https://github.com/libre-server/rolekit/issues/1

Deployed the memcache role with --next-boot. Tested it by doing 'systemctl start pendingrole-memcache-name.service'
It has not been tested with an actual kickstart yet; I am waiting for Alpha RC1 to arrive and will test it with that.