Replace multi-role checks with action plugin

This approach should make it easier to add new checks without having to
write lots of YAML and doing things against Ansible (e.g.
ignore_errors).

A single action plugin determines what checks to run per each host,
including arguments to the check. A check is implemented as a class with
a run method, with the same signature as an action plugin and module,
and is normally backed by a regular Ansible module.

Each check is implemented as a separate Python file. This allows whoever
adds a new check to focus solely in a single Python module, and
potentially an Ansible module within library/ too.

All checks are automatically loaded, and only active checks that are
requested by the playbook get executed.

21 files changed, 342 insertions, 195 deletions

diff --git a/playbooks/byo/openshift-preflight/check.yml b/playbooks/byo/openshift-preflight/check.yml
index 32673d01d..935531e02 100644
--- a/playbooks/byo/openshift-preflight/check.yml
+++ b/playbooks/byo/openshift-preflight/check.yml
@@ -1,31 +1,14 @@
 ---
 - hosts: OSEv3
-  roles:
-    - openshift_preflight/init
-
-- hosts: OSEv3
-  name: checks that apply to all hosts
-  gather_facts: no
-  ignore_errors: yes
-  roles:
-    - openshift_preflight/common
-
-- hosts: masters
-  name: checks that apply to masters
-  gather_facts: no
-  ignore_errors: yes
-  roles:
-    - openshift_preflight/masters
-
-- hosts: nodes
-  name: checks that apply to nodes
-  gather_facts: no
-  ignore_errors: yes
-  roles:
-    - openshift_preflight/nodes
-
-- hosts: OSEv3
-  name: verify check results
-  gather_facts: no
-  roles:
-    - openshift_preflight/verify_status
+  name: run OpenShift health checks
+  roles:
+    - openshift_health_checker
+  post_tasks:
+    # NOTE: we need to use the old "action: name" syntax until
+    # https://github.com/ansible/ansible/issues/20513 is fixed.
+    - action: openshift_health_check
+      args:
+        checks:
+          - package_availability
+          - package_update
+          - package_version
diff --git a/roles/openshift_health_checker/README.md b/roles/openshift_health_checker/README.md
new file mode 100644
index 000000000..745e45b50
--- /dev/null
+++ b/roles/openshift_health_checker/README.md
@@ -0,0 +1,43 @@
+OpenShift Health Checker
+========================
+
+This role detects common problems with OpenShift installations or with
+environments prior to install.
+
+Requirements
+------------
+
+* Ansible 2.2+
+
+Role Variables
+--------------
+
+None
+
+Dependencies
+------------
+
+- openshift_facts
+
+Example Playbook
+----------------
+
+```yaml
+---
+- hosts: OSEv3
+  name: run OpenShift health checks
+  roles:
+    - openshift_health_checker
+  post_tasks:
+    - action: openshift_health_check
+```
+
+License
+-------
+
+Apache License Version 2.0
+
+Author Information
+------------------
+
+Customer Success team (dev@lists.openshift.redhat.com)
diff --git a/roles/openshift_health_checker/action_plugins/openshift_health_check.py b/roles/openshift_health_checker/action_plugins/openshift_health_check.py
new file mode 100644
index 000000000..36defde0a
--- /dev/null
+++ b/roles/openshift_health_checker/action_plugins/openshift_health_check.py
@@ -0,0 +1,95 @@
+"""
+Ansible action plugin to execute health checks in OpenShift clusters.
+"""
+# pylint: disable=wrong-import-position,missing-docstring,invalid-name
+import sys
+import os
+
+try:
+    from __main__ import display
+except ImportError:
+    from ansible.utils.display import Display
+    display = Display()
+
+from ansible.plugins.action import ActionBase
+
+# Augment sys.path so that we can import checks from a directory relative to
+# this callback plugin.
+sys.path.insert(1, os.path.dirname(os.path.dirname(__file__)))
+
+from openshift_checks import OpenShiftCheck, OpenShiftCheckException  # noqa: E402
+
+
+class ActionModule(ActionBase):
+
+    def run(self, tmp=None, task_vars=None):
+        result = super(ActionModule, self).run(tmp, task_vars)
+
+        if task_vars is None:
+            task_vars = {}
+
+        if "openshift" not in task_vars:
+            result["failed"] = True
+            result["msg"] = "'openshift' is undefined, did 'openshift_facts' run?"
+            return result
+
+        try:
+            known_checks = self.load_known_checks()
+        except OpenShiftCheckException as e:
+            result["failed"] = True
+            result["msg"] = str(e)
+            return result
+
+        args = self._task.args
+        requested_checks = set(args.get("checks", []))
+
+        unknown_checks = requested_checks - set(known_checks)
+        if unknown_checks:
+            result["failed"] = True
+            result["msg"] = (
+                "One or more checks are unknown: {}. "
+                "Make sure there is no typo in the playbook and no files are missing."
+            ).format(", ".join(unknown_checks))
+            return result
+
+        result["checks"] = check_results = {}
+
+        for check_name in requested_checks & set(known_checks):
+            display.banner("CHECK [{} : {}]".format(check_name, task_vars["ansible_host"]))
+            check = known_checks[check_name]
+
+            if check.is_active(task_vars):
+                try:
+                    r = check.run(tmp, task_vars)
+                except OpenShiftCheckException as e:
+                    r = {}
+                    r["failed"] = True
+                    r["msg"] = str(e)
+            else:
+                r = {"skipped": True}
+
+            check_results[check_name] = r
+
+            if r.get("failed", False):
+                result["failed"] = True
+                result["msg"] = "One or more checks failed"
+
+        return result
+
+    def load_known_checks(self):
+        known_checks = {}
+
+        known_check_classes = set(cls for cls in OpenShiftCheck.subclasses())
+
+        for cls in known_check_classes:
+            check_name = cls.name
+            if check_name in known_checks:
+                other_cls = known_checks[check_name].__class__
+                raise OpenShiftCheckException(
+                    "non-unique check name '{}' in: '{}.{}' and '{}.{}'".format(
+                        check_name,
+                        cls.__module__, cls.__name__,
+                        other_cls.__module__, other_cls.__name__))
+            known_checks[check_name] = cls(module_executor=self._execute_module)
+
+        return known_checks
diff --git a/roles/openshift_preflight/verify_status/callback_plugins/zz_failure_summary.py b/roles/openshift_health_checker/callback_plugins/zz_failure_summary.py
index 180ed8d8f..8caefab15 100644
--- a/roles/openshift_preflight/verify_status/callback_plugins/zz_failure_summary.py
+++ b/roles/openshift_health_checker/callback_plugins/zz_failure_summary.py
@@ -3,6 +3,8 @@
 Ansible callback plugin.
 '''
 
+from pprint import pformat
+
 from ansible.plugins.callback import CallbackBase
 from ansible import constants as C
 from ansible.utils.color import stringc
@@ -79,6 +81,8 @@ def _format_failure(failure):
         (u'Task', task),
         (u'Message', stringc(msg, C.COLOR_ERROR)),
     )
+    if 'checks' in result._result:
+        rows += ((u'Details', stringc(pformat(result._result['checks']), C.COLOR_ERROR)),)
     row_format = '{:10}{}'
     return [row_format.format(header + u':', body) for header, body in rows]
 
diff --git a/roles/openshift_preflight/base/library/aos_version.py b/roles/openshift_health_checker/library/aos_version.py
index 37c8b483c..37c8b483c 100755
--- a/roles/openshift_preflight/base/library/aos_version.py
+++ b/roles/openshift_health_checker/library/aos_version.py
diff --git a/roles/openshift_preflight/base/library/check_yum_update.py b/roles/openshift_health_checker/library/check_yum_update.py
index 9bc14fd47..9bc14fd47 100755
--- a/roles/openshift_preflight/base/library/check_yum_update.py
+++ b/roles/openshift_health_checker/library/check_yum_update.py
diff --git a/roles/openshift_preflight/init/meta/main.yml b/roles/openshift_health_checker/meta/main.yml
index 0bbeadd34..0bbeadd34 100644
--- a/roles/openshift_preflight/init/meta/main.yml
+++ b/roles/openshift_health_checker/meta/main.yml
diff --git a/roles/openshift_health_checker/openshift_checks/__init__.py b/roles/openshift_health_checker/openshift_checks/__init__.py
new file mode 100644
index 000000000..d893ba591
--- /dev/null
+++ b/roles/openshift_health_checker/openshift_checks/__init__.py
@@ -0,0 +1,57 @@
+"""
+Health checks for OpenShift clusters.
+"""
+
+import os
+from abc import ABCMeta, abstractmethod, abstractproperty
+from importlib import import_module
+
+import six
+
+
+class OpenShiftCheckException(Exception):
+    """Raised when a check cannot proceed."""
+    pass
+
+
+@six.add_metaclass(ABCMeta)
+class OpenShiftCheck(object):
+    """A base class for defining checks for an OpenShift cluster environment."""
+
+    def __init__(self, module_executor):
+        self.module_executor = module_executor
+
+    @abstractproperty
+    def name(self):
+        """The name of this check, usually derived from the class name."""
+        return "openshift_check"
+
+    @classmethod
+    def is_active(cls, task_vars):  # pylint: disable=unused-argument
+        """Returns true if this check applies to the ansible-playbook run."""
+        return True
+
+    @abstractmethod
+    def run(self, tmp, task_vars):
+        """Executes a check, normally implemented as a module."""
+        return {}
+
+    @classmethod
+    def subclasses(cls):
+        """Returns a generator of subclasses of this class and its subclasses."""
+        for subclass in cls.__subclasses__():  # pylint: disable=no-member
+            yield subclass
+            for subclass in subclass.subclasses():
+                yield subclass
+
+
+# Dynamically import all submodules for the side effect of loading checks.
+
+EXCLUDES = (
+    "__init__.py",
+    "mixins.py",
+)
+
+for name in os.listdir(os.path.dirname(__file__)):
+    if name.endswith(".py") and name not in EXCLUDES:
+        import_module(__package__ + "." + name[:-3])
diff --git a/roles/openshift_health_checker/openshift_checks/mixins.py b/roles/openshift_health_checker/openshift_checks/mixins.py
new file mode 100644
index 000000000..4e0415944
--- /dev/null
+++ b/roles/openshift_health_checker/openshift_checks/mixins.py
@@ -0,0 +1,24 @@
+# pylint: disable=missing-docstring
+from openshift_checks import OpenShiftCheckException
+
+
+class NotContainerized(object):
+    """Mixin for checks that are only active when not in containerized mode."""
+
+    @classmethod
+    def is_active(cls, task_vars):
+        return (
+            # This mixin is meant to be used with subclasses of
+            # OpenShiftCheck. Pylint disables this by default on mixins,
+            # though it relies on the class name ending in 'mixin'.
+            # pylint: disable=no-member
+            super(NotContainerized, cls).is_active(task_vars) and
+            not cls.is_containerized(task_vars)
+        )
+
+    @staticmethod
+    def is_containerized(task_vars):
+        try:
+            return task_vars["openshift"]["common"]["is_containerized"]
+        except (KeyError, TypeError):
+            raise OpenShiftCheckException("'openshift.common.is_containerized' is undefined")
diff --git a/roles/openshift_health_checker/openshift_checks/package_availability.py b/roles/openshift_health_checker/openshift_checks/package_availability.py
new file mode 100644
index 000000000..4260cbf7c
--- /dev/null
+++ b/roles/openshift_health_checker/openshift_checks/package_availability.py
@@ -0,0 +1,69 @@
+# pylint: disable=missing-docstring
+from openshift_checks import OpenShiftCheck, OpenShiftCheckException
+from openshift_checks.mixins import NotContainerized
+
+
+class PackageAvailability(NotContainerized, OpenShiftCheck):
+    """Check that required RPM packages are available."""
+
+    name = "package_availability"
+
+    def run(self, tmp, task_vars):
+        try:
+            rpm_prefix = task_vars["openshift"]["common"]["service_type"]
+        except (KeyError, TypeError):
+            raise OpenShiftCheckException("'openshift.common.service_type' is undefined")
+
+        group_names = task_vars.get("group_names", [])
+
+        packages = set()
+
+        if "masters" in group_names:
+            packages.update(self.master_packages(rpm_prefix))
+        if "nodes" in group_names:
+            packages.update(self.node_packages(rpm_prefix))
+
+        args = {"packages": sorted(set(packages))}
+        return self.module_executor("check_yum_update", args, tmp, task_vars)
+
+    @staticmethod
+    def master_packages(rpm_prefix):
+        return [
+            "{rpm_prefix}".format(rpm_prefix=rpm_prefix),
+            "{rpm_prefix}-clients".format(rpm_prefix=rpm_prefix),
+            "{rpm_prefix}-master".format(rpm_prefix=rpm_prefix),
+            "bash-completion",
+            "cockpit-bridge",
+            "cockpit-docker",
+            "cockpit-kubernetes",
+            "cockpit-shell",
+            "cockpit-ws",
+            "etcd",
+            "httpd-tools",
+        ]
+
+    @staticmethod
+    def node_packages(rpm_prefix):
+        return [
+            "{rpm_prefix}".format(rpm_prefix=rpm_prefix),
+            "{rpm_prefix}-node".format(rpm_prefix=rpm_prefix),
+            "{rpm_prefix}-sdn-ovs".format(rpm_prefix=rpm_prefix),
+            "bind",
+            "ceph-common",
+            "dnsmasq",
+            "docker",
+            "firewalld",
+            "flannel",
+            "glusterfs-fuse",
+            "iptables-services",
+            "iptables",
+            "iscsi-initiator-utils",
+            "libselinux-python",
+            "nfs-utils",
+            "ntp",
+            "openssl",
+            "pyparted",
+            "python-httplib2",
+            "PyYAML",
+            "yum-utils",
+        ]
diff --git a/roles/openshift_health_checker/openshift_checks/package_update.py b/roles/openshift_health_checker/openshift_checks/package_update.py
new file mode 100644
index 000000000..316a776f5
--- /dev/null
+++ b/roles/openshift_health_checker/openshift_checks/package_update.py
@@ -0,0 +1,13 @@
+# pylint: disable=missing-docstring
+from openshift_checks import OpenShiftCheck
+from openshift_checks.mixins import NotContainerized
+
+
+class PackageUpdate(NotContainerized, OpenShiftCheck):
+    """Check that there are no conflicts in RPM packages."""
+
+    name = "package_update"
+
+    def run(self, tmp, task_vars):
+        args = {"packages": []}
+        return self.module_executor("check_yum_update", args, tmp, task_vars)
diff --git a/roles/openshift_health_checker/openshift_checks/package_version.py b/roles/openshift_health_checker/openshift_checks/package_version.py
new file mode 100644
index 000000000..a473119f3
--- /dev/null
+++ b/roles/openshift_health_checker/openshift_checks/package_version.py
@@ -0,0 +1,25 @@
+# pylint: disable=missing-docstring
+from openshift_checks import OpenShiftCheck, OpenShiftCheckException
+from openshift_checks.mixins import NotContainerized
+
+
+class PackageVersion(NotContainerized, OpenShiftCheck):
+    """Check that available RPM packages match the required versions."""
+
+    name = "package_version"
+
+    @classmethod
+    def is_active(cls, task_vars):
+        return (
+            super(PackageVersion, cls).is_active(task_vars)
+            and task_vars.get("deployment_type") == "openshift-enterprise"
+        )
+
+    def run(self, tmp, task_vars):
+        try:
+            openshift_release = task_vars["openshift_release"]
+        except (KeyError, TypeError):
+            raise OpenShiftCheckException("'openshift_release' is undefined")
+
+        args = {"version": openshift_release}
+        return self.module_executor("aos_version", args, tmp, task_vars)
diff --git a/roles/openshift_preflight/README.md b/roles/openshift_preflight/README.md
deleted file mode 100644
index b6d3542d3..000000000
--- a/roles/openshift_preflight/README.md
+++ /dev/null
@@ -1,52 +0,0 @@
-OpenShift Preflight Checks
-==========================
-
-This role detects common problems prior to installing OpenShift.
-
-Requirements
-------------
-
-* Ansible 2.2+
-
-Role Variables
---------------
-
-None
-
-Dependencies
-------------
-
-None
-
-Example Playbook
-----------------
-
-```yaml
----
-- hosts: OSEv3
-  roles:
-    - openshift_preflight/init
-
-- hosts: OSEv3
-  name: checks that apply to all hosts
-  gather_facts: no
-  ignore_errors: yes
-  roles:
-    - openshift_preflight/common
-
-- hosts: OSEv3
-  name: verify check results
-  gather_facts: no
-  roles:
-    - openshift_preflight/verify_status
-```
-
-License
--------
-
-Apache License Version 2.0
-
-Author Information
-------------------
-
-Customer Success team (dev@lists.openshift.redhat.com)
diff --git a/roles/openshift_preflight/common/meta/main.yml b/roles/openshift_preflight/common/meta/main.yml
deleted file mode 100644
index 6f23cbf3b..000000000
--- a/roles/openshift_preflight/common/meta/main.yml
+++ /dev/null
@@ -1,3 +0,0 @@
----
-dependencies:
-  - role: openshift_preflight/base
diff --git a/roles/openshift_preflight/common/tasks/main.yml b/roles/openshift_preflight/common/tasks/main.yml
deleted file mode 100644
index f1a4a160e..000000000
--- a/roles/openshift_preflight/common/tasks/main.yml
+++ /dev/null
@@ -1,21 +0,0 @@
----
-# check content available on all hosts
-- when: not openshift.common.is_containerized | bool
-  block:
-
-    - name: determine if yum update will work
-      action: check_yum_update
-      register: r
-
-    - set_fact:
-        oo_preflight_check_results: "{{ oo_preflight_check_results + [r|combine({'_task': 'determine if yum update will work'})] }}"
-
-    - name: determine if expected version matches what is available
-      aos_version:
-        version: "{{ openshift_release }}"
-      when:
-        - deployment_type == "openshift-enterprise"
-      register: r
-
-    - set_fact:
-        oo_preflight_check_results: "{{ oo_preflight_check_results + [r|combine({'_task': 'determine if expected version matches what is available'})] }}"
diff --git a/roles/openshift_preflight/init/tasks/main.yml b/roles/openshift_preflight/init/tasks/main.yml
deleted file mode 100644
index bf2d82196..000000000
--- a/roles/openshift_preflight/init/tasks/main.yml
+++ /dev/null
@@ -1,4 +0,0 @@
----
-- name: set common variables
-  set_fact:
-    oo_preflight_check_results: "{{ oo_preflight_check_results | default([]) }}"
diff --git a/roles/openshift_preflight/masters/meta/main.yml b/roles/openshift_preflight/masters/meta/main.yml
deleted file mode 100644
index 6f23cbf3b..000000000
--- a/roles/openshift_preflight/masters/meta/main.yml
+++ /dev/null
@@ -1,3 +0,0 @@
----
-dependencies:
-  - role: openshift_preflight/base
diff --git a/roles/openshift_preflight/masters/tasks/main.yml b/roles/openshift_preflight/masters/tasks/main.yml
deleted file mode 100644
index 35fb1e3ca..000000000
--- a/roles/openshift_preflight/masters/tasks/main.yml
+++ /dev/null
@@ -1,31 +0,0 @@
----
-# determine if yum install of master pkgs will work
-- when: not openshift.common.is_containerized | bool
-  block:
-
-    - name: main master packages availability
-      check_yum_update:
-        packages:
-          - "{{ openshift.common.service_type }}"
-          - "{{ openshift.common.service_type }}-clients"
-          - "{{ openshift.common.service_type }}-master"
-      register: r
-
-    - set_fact:
-        oo_preflight_check_results: "{{ oo_preflight_check_results + [r|combine({'_task': 'main master packages availability'})] }}"
-
-    - name: other master packages availability
-      check_yum_update:
-        packages:
-          - etcd
-          - bash-completion
-          - cockpit-bridge
-          - cockpit-docker
-          - cockpit-kubernetes
-          - cockpit-shell
-          - cockpit-ws
-          - httpd-tools
-      register: r
-
-    - set_fact:
-        oo_preflight_check_results: "{{ oo_preflight_check_results + [r|combine({'_task': 'other master packages availability'})] }}"
diff --git a/roles/openshift_preflight/nodes/meta/main.yml b/roles/openshift_preflight/nodes/meta/main.yml
deleted file mode 100644
index 6f23cbf3b..000000000
--- a/roles/openshift_preflight/nodes/meta/main.yml
+++ /dev/null
@@ -1,3 +0,0 @@
----
-dependencies:
-  - role: openshift_preflight/base
diff --git a/roles/openshift_preflight/nodes/tasks/main.yml b/roles/openshift_preflight/nodes/tasks/main.yml
deleted file mode 100644
index a10e69024..000000000
--- a/roles/openshift_preflight/nodes/tasks/main.yml
+++ /dev/null
@@ -1,41 +0,0 @@
----
-# determine if yum install of node pkgs will work
-- when: not openshift.common.is_containerized | bool
-  block:
-
-    - name: main node packages availability
-      check_yum_update:
-        packages:
-          - "{{ openshift.common.service_type }}"
-          - "{{ openshift.common.service_type }}-node"
-          - "{{ openshift.common.service_type }}-sdn-ovs"
-      register: r
-
-    - set_fact:
-        oo_preflight_check_results: "{{ oo_preflight_check_results + [r|combine({'_task': 'main node packages availability'})] }}"
-
-    - name: other node packages availability
-      check_yum_update:
-        packages:
-          - docker
-          - PyYAML
-          - firewalld
-          - iptables
-          - iptables-services
-          - nfs-utils
-          - ntp
-          - yum-utils
-          - dnsmasq
-          - libselinux-python
-          - ceph-common
-          - glusterfs-fuse
-          - iscsi-initiator-utils
-          - pyparted
-          - python-httplib2
-          - openssl
-          - flannel
-          - bind
-      register: r
-
-    - set_fact:
-        oo_preflight_check_results: "{{ oo_preflight_check_results + [r|combine({'_task': 'other node packages availability'})] }}"
diff --git a/roles/openshift_preflight/verify_status/tasks/main.yml b/roles/openshift_preflight/verify_status/tasks/main.yml
deleted file mode 100644
index 36ccf648a..000000000
--- a/roles/openshift_preflight/verify_status/tasks/main.yml
+++ /dev/null
@@ -1,8 +0,0 @@
----
-- name: find check failures
-  set_fact:
-    oo_preflight_check_failures: "{{ oo_preflight_check_results | select('failed', 'equalto', True) | list }}"
-
-- name: ensure all checks succeed
-  action: fail
-  when: oo_preflight_check_failures