Fixed a false negative related to replication taking some time: added
wait_for_replication call before checking for new object in replicas.
Reviewed-By: Martin Babinsky <mbabinsk@redhat.com>
The following testcases were automated:
1. Test one command replica installation
2. Test csreplica-manage-(del, connect, disconnect) are disabled in domain
level 1
3. Client enrollment and replica promotion by an unprivileged user are
prohibited
4. Replica uninstallation is prohibited if it disconnects a part of existing
topology (is possible only with --ignore-topology-disconnect option)
https://fedorahosted.org/freeipa/ticket/5723
Reviewed-By: Martin Babinsky <mbabinsk@redhat.com>
Some integration tests expect the members of topology to have the testfolder.
The testfolder however was created during master and client installations. In
case the client is not intalled from the very beginning of the test execution,
the test would complain about missing testfolder.
https://fedorahosted.org/freeipa/ticket/5723
Reviewed-By: Martin Basti <mbasti@redhat.com>
teardown_method is a standard pytest method used to put any code to be executed
after each test method is executed. While treated correctly by our integration
tests, this method is misinterpreted by in-tree tests in the following way:
in-tree tests try to execute it even if all the test methods are skipped due to
test resources being not configured. This causes the tests, that otherwise would
have been skipped, to fail
https://fedorahosted.org/freeipa/ticket/5723
Reviewed-By: Martin Basti <mbasti@redhat.com>
prepare_host is executed from within each of install_master, install_replica
and install_client in tasks.py anyway, so no need to call it here also.
Besindes this call kept failing when IntegrationTest wes initialized more than
once during the test execution.
https://fedorahosted.org/freeipa/ticket/5723
Reviewed-By: Martin Basti <mbasti@redhat.com>
Explicitly specifying ip-address of the replica messes up with the current
bind-dyndb-ldap logic, causing reverse zone not to be created.
Enabled reverse-zone creation for the clients residing in different subnet from
master
Reviewed-By: Petr Spacek <pspacek@redhat.com>
Without it any test comprized of more than one cycle of installing-uninstalling
of ipa would fail due to the fact that test folder on the remote machine gets
deleted during ipa uninstallation.
Also removed duplicate call of apply_common fixes and added unapply_fixes to
uninstall_replica
Reviewed-By: Martin Basti <mbasti@redhat.com>
When resolv.conf is set to point to the master's ip before installation, the
ipa-server-install does not create a reverse zone for it's ip even despite
--auto-reverse option provided. The fix is not to mess around with resolv.conf
before master installation.
Reviewed-By: Petr Spacek <pspacek@redhat.com>
Without realm provided explicitly, installation calculates it automatically
from the current hostname which may be inconsistent with the configured domain
name. Which, in turn, causes failures in integration tests in the lab.
Reviewed-By: Martin Basti <mbasti@redhat.com>
As of 4.3 the replica installation is performed without preparing a gpg file on
master, but rather enrolling a future replica as a client with subsequent
promotion of the client. This required the corresponding change in the
integration tests
https://fedorahosted.org/freeipa/ticket/5379
Reviewed-By: Martin Basti <mbasti@redhat.com>
In beaker lab the situation when master and replica have ip addresses from
different subnets is quite frequent. When a replica has ip from different
subnet than master's, ipa-replica-prepare looks up a proper reverse zone to
add a pointer record, and if it does not find it, it asks a user for permission
to create it automatically. It breaks the tests adding the unexpected input.
The workaround is to always create a reverse zone for a new replica.
Corresponding ticket is https://fedorahosted.org/freeipa/ticket/5306
Reviewed-By: Petr Spacek <pspacek@redhat.com>
Reviewed-By: Martin Basti <mbasti@redhat.com>
Oleg Fayans