As the old way to include tasks is deprecated, replace static include
statements with import_tasks and dynamic ones with include_tasks.
Increaded the required ansible version to 2.5.0 to make sure that
include_tasks and import_tasks is working as expected.
Fixes issue #38
These roles will most likely not work in the common case. Therefore the roles
have been renamed.
The ipa-krpb5 role is used by ipcalient, but the ipa-sssd role is currently
not used.
As the action plugin is used with the default python interpreter and
the change to python3 for FreeIPA, the use of OTP was not working anymore.
The ansible_python_interpreter is not automatically used for the module
part of the action plugin. Therefore ansible_python_interpreter needed to
be added to the action plugin call as a new var to make sure that the
module part is used with the proper python version.
Also a new import for the Python2/3 import test has been added to discover
of the server is supporting python2 or python3. The old
ansible_python_interpreter setting is saved before doing this and restored
after the one-time password has been generated on the server.
With IPA 4.7 bigger changes have been introduced
Changes:
- Use of timeconf and chrony instead of ntpconf and ntpd.
- A new option ntp_pool has been introduced.
krb5 DNS discovery was not possible in cluster environments as the server
list from groups.ipaserver was used all the time. DNS discovery is though
only used if no servers are given.
The new setting ipaclient_no_dns_lookup has been added to make sure that
DNS lookup is used in the first place and can be disabled easily with this
setting. There is also a new way to override servers per client in the
inventory file with ipaclient_servers.
Two new settings have been added:
ipaclient_no_dns_lookup (bool, default: no)
Set to 'yes' to use groups.ipaserver in cluster environments as servers
for the clients. This deactivates DNS lookup in krb5.
ipaclient_servers (list of strings, default: undefined)
Manually override list of servers for example in a cluster environment on
a per client basis. The list of servers is normally taken from from
groups.ipaserver in cluster environments.
The krb5 DNS lookup settings krb5_dns_lookup_realm and krb5_dns_lookup_kdc
ans also the servers have not been set properly set if no server has been
specified and discovery succeeded. This has been fixed.
This fixes issue #23.
- Do not register a change in the playbook run when registering the
variable checking for whether or not Python 3 imports work
Signed-off-by: Kellin <kellin@retromud.org>
The client role is used also while installing the server. There has been an
issue where the server installation has not been complete because of a
playbook termination in the client.
This has been fixed and the client and also the server are fully configured
in the server installation.
The relative import of the distribution specific vars files requires to use
is not working. {{ role_path }} needs to be used to force the load of the
proper files.
The server role has different setting names:
- groups.ipaserver: groups.ipaservers
- ipaserver_domain: ipaclient_domain
- ipaserver_realm: ipaclient_realm
Both need to be supported to be able to sue the client role within the server
role, but also standalone.
Attempt to sync time if on_master is not set and no_ntp is not set: At
first with given or dicovered time servers. If no ntp servers have been
given or discovered, then with the ipa server.
New parameters:
on_master:
description: IPA client installation on IPA server
required: false
default: false
type: bool
default: no
ntp_servers:
description: List of NTP servers to use
required: false
type: list
default: []
no_ntp:
description: Do not sync time and do not detect time servers
required: false
default: false
type: bool
default: no
The ntp_servers output parameter is now always an empty list if on_master
or no_ntp is set.
ipaclient_hostname needs to be specified in the inventory file for the hosts
where the name needs to get changed.
Example:
192.168.1.1 ipaclient_hostname=ipaclient1.mine.local
The option should not be specified in [ipaclients:vars] as all hosts would
get the same name.
Currently ipaclient role is using the module ipaclient only for uninstallation,
and this module contains a lot of unused code.
It is simpler to directly call the command-line
ipa-client-install --uninstall -U
and remove the ipaclient module.
When the client already has a working keytab, use_otp is disabled. This creates
an issue when ipaclient_force_join is set, because the join module is called
with ipaadmin_principal and ipaadmin_password, but these variables may be
undefined if ipaadmin_keytab is used instead.
We should not disable OTP when force-join is specified.
With the test it is not needed to pin down the python interpreter for ansible
modules. It is therefore possible to use a Python2 version on Fedora-27 and
a Python3 version on Fedora-26.
In the client krb5.conf setup, a pkinit_anchors entry
was being added for pki-ca-bundle. This should instead
be kdc-ca-bundle.
Signed-off-by: Scott Poore <spoore@redhat.com>
The new results from ipatest (krb5_conf_ok and ipa_test_ok) are now used for
additional fails to suggest to enable allow_repair.
The playbook is not ended anymore if ipajoin changed something.
The rename was needed to be able to have more than one package in the list
of required packages.
For RHEL-7.3 it has been needed to add ipa-admintools to have /usr/bin/ipa
available. libselinux-python has been added for all.
Add big block has been added that contains all steps where the ccache is
created an used. With the block it is possible to add an always clause to
remove the ccachae also in the error case. The cleanup of the ccache is
also done in the beginning to make sure that no ccache leftover will be
used.
The subject base generated in discovery is only a guess and might have been
changed by the admin at installation process. Therefore it is needed to
get this from the server.
subject_base has been added as a new return value.
Use subject base form ipaapi in roles/ipaclient/tasks/install.yml instead of
guessed value from ipadiscovery.
When allow_repair is enabled, then the playbook will continue for an
already joined host. The remaining steps ipaconf, ipasssd, krb5, ipaapi,
ipanss and ipaextras will be redone.
If allow_repair is disabled, then the meta module will be
used with the end_play option to stop the processing of the playbook
without an error.
If a working krb5.keytab has been detected on the host then use_otp will be
disabled and join will not be called. This is done to preserve the keytab
entry in the host entry on the server.
Enforcing the creation of a one-time-password will result in a host-disable
call for the host entry. This will remove an existing keytab and password from
the entry.
