The raises of RuntimeError, ValueError and ScriptError are currently not
properly handled in ipaserver_prepare. This results in a trace back error
shown in Ansible instead of only showing the error message.
This happened for example if a nameserver is in /etc/resolv.conf that is
not reachable.
The use of password will conflict with the user password setting and is
not really descriptive. ipaadmin_principal and ipaadmin_password are also
used in the roles.
This adds support for the --external-ca option to ipaserver. Lots of
additional tests and checks from ServerInstallInterface.__init__ have
been added to ipaserver_test. Also duplicate tests cna checks have been
removed.
Installer settings in ansible_ipa_server module_util are now also set
to the defaults that are used in Installable, ServerInstallInterface,
ServerMasterInstall, ADTrustInstallInterface and Uninstall.
The /root/ipa.csr file generated on the node in ca.install_step_0 will
be copied to the controller as "{{ inventory_hostname }}-ipa.csr".
The new task file copy_external_cert.yml has been added to copy the
generated certificate defined in ipaserver_external_cert_files to the node
to continue with ca.install_step_1.
The tasks/install.yml file has been adapted to make sure that the steps
that will be done in step two will be skipped after step one has been
done.
These two settings are not set using the krb.init_info method, but used in
krb.enable_ssl.
The configuration of PKINIT fails in IPA 4.7.0 because of the issue
https://pagure.io/freeipa/issue/7655 where auto detection of the
master is not properly working. With the missing setting of krb.master_fqdn
the not workint auto detection has been triggered, which resulted in
failed PKINIT enablement.
Some settings for kra have not been correct for kra with the change to
use single Custodia instance in the installer (freeipa 994f71ac8).
These modules have been adapted:
ipareplica_custodia_import_dm_password
ipareplica_enable_ipa
ipareplica_setup_ca
ipareplica_setup_custodia
ipareplica_setup_kra
This is related to freeipa#0f31564b35aac250456233f98730811560eda664
During ipa-replica-install, http installation first creates a service
principal for http/hostname (locally on the soon-to-be-replica), then
waits for this entry to be replicated on the master picked for the
install.
In a later step, the installer requests a certificate for HTTPd. The local
certmonger first tries the master defined in xmlrpc_uri (which is
pointing to the soon-to-be-replica), but fails because the service is not
up yet. Then certmonger tries to find a master by using the DNS and looking
for a ldap service. This step can pick a different master, where the
principal entry has not always be replicated yet.
As the certificate request adds the principal if it does not exist, we can
end by re-creating the principal and have a replication conflict.
The replication conflict later causes kerberos issues, preventing
from installing a new replica.
The proposed fix forces xmlrpc_uri to point to the same master as the one
picked for the installation, in order to make sure that the master already
contains the principal entry.
https://pagure.io/freeipa/issue/7041
This has not been done so far in the ansible_ipa_client, but only in the
modules where it was really needed. But as these places are getting more
with 4.7.90, this setting makes it into the module_utils.
It is now possible to use domain+ca as suffix, That means that the segment
will be handled for the suffixes domain and also ca.
The new state checked is returning two lists found and not-found. If a
segment exists, the ckecked suffix is added to the found list. If a segment
from suffix is not found, it is added to the not-found list.
New example playbooks have been added:
playbooks/topology/add-topologysegments.yml
playbooks/topology/check-topologysegments.yml
playbooks/topology/delete-topologysegments.yml
The cluster playbook has been extended by the
The module ipareplica_master_password has been a copy from ipaserver role
and still contained code to read the cache file. This is not needed for
the replica. Therefore there is no need also to provide the dm password
to ipareplica_master_password any more.
Currently the error message of ipaclient_get_otp is not visble if it failed
due to the use of no_log. The no_log option is needed and useful if the
one-time password has successfully been generated, but is bad if there was
an error while generating the password, for example if a clock skew has been
detected.
There is now a new task to print the error message if ipaclient_get_otp. The
task for ipaclient_get_otp has been marked with "ignore_errors: yes" and
the new task "Install - Report error for OTP generation" will always fail
and only be used if result_ipaclient_get_otp is failed.
ipaadmin_keytab has been supported only with with ipaclient_use_otp. But
it can also be used without for ipa-join.
Important is that ipaadmin_keytab needs to be placed on the cliend node
and ipaadmin_keytab needs to be a full path. Otherwise the file will not
be found.
Password and keytab do not need to be set explicitely to an empty string
when they are not set. Also there is no need to have string length checks
in the role tasks.
With the deactivation of the Python2/3 test the handling of
ansible_python_interpreter needs to be removed as the setting might
not exist and is not changed in with the Python2/3 test any more.
The code for host_name, the domain_name and also the realm_name has been
adapted to the code in the command line installer. The _hostname_overridden
setting is now only true if the hostname has been changed.
The install checks have been done temporarily in _test and finally also
in _prepare. This is not needed and also not done this way in the command
line installers.
The addtion is not oly adding the config setting, but also fixing the
deployment without the setting as functions and methods have been changed
for pki_config_override.
There is a new setting for the ipaserver role:
ipaserver_pki_config_override
tasks.restore_context is only used in old releases. The existence of
paths.CACHE_IPA_SESSIONS is used to determine if the call needs to be
done or not.
There are now two topology management modules placed in the plugins folder:
plugins/modules/ipatopologysegment.py
plugins/modules/ipatopologysuffix.py
Topology segments can be added, removed and reinitialized with the
ipatopologysegment module. Also it is possible to verify topology suffixes
with the ipatopologysuffix module.
A new module_utils for plugins has been added:
plugins/module_utils/ansible_freeipa_module.py
And documentation for the modules:
README-topology.md
New sample playbooks are available in playbooks/topology:
playbooks/topology/add-topologysegment.yml
playbooks/topology/delete-topologysegment.yml
playbooks/topology/reinitialize-topologysegment.yml
playbooks/topology/verify-topologysuffix.yml
The plugins folder can be used with the new Ansible Collections supported
by Ansible 2.8 and Ansible galaxy 3.2.
The playbooks install-client.yml, install-cluster.yml, install-replica.yml,
install-server.yml, uninstall-client.yml, uninstall-cluster.yml,
uninstall-replica.yml and uninstall-server.yml have been moved into
the playbooks folder.
