group_remove_member is not able to handle services in old IPA releases.
In one case the check was missing and the removal of a user from a group
failed because of this with an older IPA version. The missing check has
been added.
Fixes#257 (ipagroup fails to remove user from group ipausers)
The host_find command had to be replaced to get the "has_password" and
"has_keytab" return values. This commit replaces the dnsrecord_find
with the dnsrecord_show command to have consistent find functions in
the module.
If random is enabled and update_password is limited to "create_only", the
random password may only be changed if the host does not exist yet.
Additionally the generation of the random password will fail, if the host
is already enrolled if update_password is "always" (default value). An
error will be reported early in this case now.
The command host_show is now used instead of host_find, as `has_password`
and `has_keytab` are only returned by host_show, but not by host_find. The
find_host function has been adapated for this change.
Resolves: #253 (ipahost is not idempotent)
The find_user function was not using encode_certificate for certificates
that are stored in the user record. This could lead to some issues with
older ipa releases and Python 2.
Because of a missing check member attributes (for use with action: member)
are cleared when a non-member attribute is changed. The fix simply adds a
check for None (parameter not set) to gen_add_del_lists in
ansible_freeipa_module to make sure that the parameter is only changed if
it should be changed.
All places where the add and removal lists have been generated manually
have been changed to also use gen_add_del_lists.
Resolves: #252 (The "Manager" attribute is removed when updating any user
attribute)
Even after obtaining Kerberos TGT with temp_kinit(), when connecting to
the IPA API with context `ansible-freeipa`, the API commands complained
that Kerberos credentials were not available. This patch fixes this
behavior.
In the case that the admin password has been set and become was not set
the call to backend.connect in api_connect failed. The solution is simply
to set os.environ["KRB5CCNAME"] in temp_kinit after kinit_password has
been called using the temporary ccache. os.environ["KRB5CCNAME"] is not
used automatically by api.Backend.[ldap2,rpcclient].connect. Afterwards
os.environ["KRB5CCNAME"] is unset in temp_kdestroy if ccache_name is not
None.
Fixes: #249 (Kerberos errors while using the modules with a non-sudoer user)
This patch allows the removal of option `all` from user, host, group,
runasuser, and runasgroup categories, by allowing an empty string as
a valid choice for each option.
The namespace and colleciton name have been hard coded. Now variables are
used for them. The project prefix and collection prefix are now passed to
galaxyify-playbook.py.
The use of "default: idstart+199999" in the description of the idmax
parameter was resulting in the galaxy import error:
Cannot parse "DOCUMENTATION": mapping values are not allowed here in
"<unicode string>", line 52, column 58: ... value for the IDs range
(default: idstart+199999)
The ":" has simply been removed to fix this issue.
When accessing an instance of AnsibleFreeIPAParams with .get the obj was
by-passing the call to _afm_convert which was the primaty reason why it
was created.
Also the class now extends Mapping instead of dict.
There is a new management module placed in the plugins folder:
plugins/modules/ipadnszone.py
The dnszone module allows to manage DNS zones.
Here is the documentation for the module:
README-dnszone.md
New example playbooks have been added:
playbooks/dnszone/disable-zone-forwarders.yml
playbooks/dnszone/dnszone-absent.yml
playbooks/dnszone/dnszone-all-params.yml
playbooks/dnszone/dnszone-disable.yml
playbooks/dnszone/dnszone-enable.yml
playbooks/dnszone/dnszone-present.yml
New tests for the module:
tests/dnszone/test_dnszone.yml
tests/dnszone/test_dnszone_mod.yml
The import of ansible_ipa_server, ansible_ipa_replica and ansible_ipa_client
might result in a permission denied error for the log file. It seems that
for collections the module utils seem to be loaded before the needed
permissions are aquired now.
The fix simply adds a wrapper for standard_logging_setup that is called in
all the modules of the server, replica and client roles to do the loggin
setup as one of the first steps of the module execution and not before.
