diff options
Diffstat (limited to 'roles/openshift_health_checker')
44 files changed, 4960 insertions, 284 deletions
diff --git a/roles/openshift_health_checker/action_plugins/openshift_health_check.py b/roles/openshift_health_checker/action_plugins/openshift_health_check.py index cf0fe19f1..0390dc82e 100644 --- a/roles/openshift_health_checker/action_plugins/openshift_health_check.py +++ b/roles/openshift_health_checker/action_plugins/openshift_health_check.py @@ -4,6 +4,7 @@ Ansible action plugin to execute health checks in OpenShift clusters. # pylint: disable=wrong-import-position,missing-docstring,invalid-name import sys import os +from collections import defaultdict try: from __main__ import display @@ -24,9 +25,11 @@ class ActionModule(ActionBase): def run(self, tmp=None, task_vars=None): result = super(ActionModule, self).run(tmp, task_vars) + task_vars = task_vars or {} - if task_vars is None: - task_vars = {} + # vars are not supportably available in the callback plugin, + # so record any it will need in the result. + result['playbook_context'] = task_vars.get('r_openshift_health_checker_playbook_context') if "openshift" not in task_vars: result["failed"] = True @@ -35,38 +38,36 @@ class ActionModule(ActionBase): try: known_checks = self.load_known_checks() + args = self._task.args + resolved_checks = resolve_checks(args.get("checks", []), known_checks.values()) except OpenShiftCheckException as e: result["failed"] = True result["msg"] = str(e) return result - args = self._task.args - requested_checks = resolve_checks(args.get("checks", []), known_checks.values()) - - 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): + user_disabled_checks = [ + check.strip() + for check in task_vars.get("openshift_disable_check", "").split(",") + ] + + for check_name in resolved_checks: display.banner("CHECK [{} : {}]".format(check_name, task_vars["ansible_host"])) check = known_checks[check_name] - if check.is_active(task_vars): + if not check.is_active(task_vars): + r = dict(skipped=True, skipped_reason="Not active for this host") + elif check_name in user_disabled_checks: + r = dict(skipped=True, skipped_reason="Disabled by user request") + else: try: r = check.run(tmp, task_vars) except OpenShiftCheckException as e: - r = {} - r["failed"] = True - r["msg"] = str(e) - else: - r = {"skipped": True} + r = dict( + failed=True, + msg=str(e), + ) check_results[check_name] = r @@ -81,10 +82,7 @@ class ActionModule(ActionBase): load_checks() known_checks = {} - - known_check_classes = set(cls for cls in OpenShiftCheck.subclasses()) - - for cls in known_check_classes: + for cls in OpenShiftCheck.subclasses(): check_name = cls.name if check_name in known_checks: other_cls = known_checks[check_name].__class__ @@ -94,26 +92,45 @@ class ActionModule(ActionBase): cls.__module__, cls.__name__, other_cls.__module__, other_cls.__name__)) known_checks[check_name] = cls(execute_module=self._execute_module) - return known_checks def resolve_checks(names, all_checks): """Returns a set of resolved check names. - Resolving a check name involves expanding tag references (e.g., '@tag') with - all the checks that contain the given tag. + Resolving a check name expands tag references (e.g., "@tag") to all the + checks that contain the given tag. OpenShiftCheckException is raised if + names contains an unknown check or tag name. names should be a sequence of strings. all_checks should be a sequence of check classes/instances. """ - resolved = set() - for name in names: - if name.startswith("@"): - for check in all_checks: - if name[1:] in check.tags: - resolved.add(check.name) - else: - resolved.add(name) + known_check_names = set(check.name for check in all_checks) + known_tag_names = set(name for check in all_checks for name in check.tags) + + check_names = set(name for name in names if not name.startswith('@')) + tag_names = set(name[1:] for name in names if name.startswith('@')) + + unknown_check_names = check_names - known_check_names + unknown_tag_names = tag_names - known_tag_names + + if unknown_check_names or unknown_tag_names: + msg = [] + if unknown_check_names: + msg.append('Unknown check names: {}.'.format(', '.join(sorted(unknown_check_names)))) + if unknown_tag_names: + msg.append('Unknown tag names: {}.'.format(', '.join(sorted(unknown_tag_names)))) + msg.append('Make sure there is no typo in the playbook and no files are missing.') + raise OpenShiftCheckException('\n'.join(msg)) + + tag_to_checks = defaultdict(set) + for check in all_checks: + for tag in check.tags: + tag_to_checks[tag].add(check.name) + + resolved = check_names.copy() + for tag in tag_names: + resolved.update(tag_to_checks[tag]) + return resolved diff --git a/roles/openshift_health_checker/callback_plugins/zz_failure_summary.py b/roles/openshift_health_checker/callback_plugins/zz_failure_summary.py index 208e81048..64c29a8d9 100644 --- a/roles/openshift_health_checker/callback_plugins/zz_failure_summary.py +++ b/roles/openshift_health_checker/callback_plugins/zz_failure_summary.py @@ -1,8 +1,13 @@ -# vim: expandtab:tabstop=4:shiftwidth=4 ''' Ansible callback plugin. ''' +# Reason: In several locations below we disable pylint protected-access +# for Ansible objects that do not give us any public way +# to access the full details we need to report check failures. +# Status: disabled permanently or until Ansible object has a public API. +# This does leave the code more likely to be broken by future Ansible changes. + from pprint import pformat from ansible.plugins.callback import CallbackBase @@ -21,38 +26,37 @@ class CallbackModule(CallbackBase): CALLBACK_TYPE = 'aggregate' CALLBACK_NAME = 'failure_summary' CALLBACK_NEEDS_WHITELIST = False + _playbook_file = None def __init__(self): super(CallbackModule, self).__init__() self.__failures = [] + def v2_playbook_on_start(self, playbook): + super(CallbackModule, self).v2_playbook_on_start(playbook) + # re: playbook attrs see top comment # pylint: disable=protected-access + self._playbook_file = playbook._file_name + def v2_runner_on_failed(self, result, ignore_errors=False): super(CallbackModule, self).v2_runner_on_failed(result, ignore_errors) self.__failures.append(dict(result=result, ignore_errors=ignore_errors)) def v2_playbook_on_stats(self, stats): super(CallbackModule, self).v2_playbook_on_stats(stats) - # TODO: update condition to consider a host var or env var to - # enable/disable the summary, so that we can control the output from a - # play. if self.__failures: - self._print_failure_summary() + self._print_failure_details(self.__failures) - def _print_failure_summary(self): - '''Print a summary of failed tasks (including ignored failures).''' + def _print_failure_details(self, failures): + '''Print a summary of failed tasks or checks.''' self._display.display(u'\nFailure summary:\n') - # TODO: group failures by host or by task. If grouped by host, it is - # easy to see all problems of a given host. If grouped by task, it is - # easy to see what hosts needs the same fix. - - width = len(str(len(self.__failures))) + width = len(str(len(failures))) initial_indent_format = u' {{:>{width}}}. '.format(width=width) initial_indent_len = len(initial_indent_format.format(0)) subsequent_indent = u' ' * initial_indent_len subsequent_extra_indent = u' ' * (initial_indent_len + 10) - for i, failure in enumerate(self.__failures, 1): + for i, failure in enumerate(failures, 1): entries = _format_failure(failure) self._display.display(u'\n{}{}'.format(initial_indent_format.format(i), entries[0])) for entry in entries[1:]: @@ -60,11 +64,52 @@ class CallbackModule(CallbackBase): indented = u'{}{}'.format(subsequent_indent, entry) self._display.display(indented) - -# Reason: disable pylint protected-access because we need to access _* -# attributes of a task result to implement this method. -# Status: permanently disabled unless Ansible's API changes. -# pylint: disable=protected-access + failed_checks = set() + playbook_context = None + # re: result attrs see top comment # pylint: disable=protected-access + for failure in failures: + # get context from check task result since callback plugins cannot access task vars + playbook_context = playbook_context or failure['result']._result.get('playbook_context') + failed_checks.update( + name + for name, result in failure['result']._result.get('checks', {}).items() + if result.get('failed') + ) + if failed_checks: + self._print_check_failure_summary(failed_checks, playbook_context) + + def _print_check_failure_summary(self, failed_checks, context): + checks = ','.join(sorted(failed_checks)) + # NOTE: context is not set if all failures occurred prior to checks task + summary = ( + '\n' + 'The execution of "{playbook}"\n' + 'includes checks designed to fail early if the requirements\n' + 'of the playbook are not met. One or more of these checks\n' + 'failed. To disregard these results, you may choose to\n' + 'disable failing checks by setting an Ansible variable:\n\n' + ' openshift_disable_check={checks}\n\n' + 'Failing check names are shown in the failure details above.\n' + 'Some checks may be configurable by variables if your requirements\n' + 'are different from the defaults; consult check documentation.\n' + 'Variables can be set in the inventory or passed on the\n' + 'command line using the -e flag to ansible-playbook.\n' + ).format(playbook=self._playbook_file, checks=checks) + if context in ['pre-install', 'health']: + summary = ( + '\n' + 'You may choose to configure or disable failing checks by\n' + 'setting Ansible variables. To disable those above:\n\n' + ' openshift_disable_check={checks}\n\n' + 'Consult check documentation for configurable variables.\n' + 'Variables can be set in the inventory or passed on the\n' + 'command line using the -e flag to ansible-playbook.\n' + ).format(checks=checks) + # other expected contexts: install, upgrade + self._display.display(summary) + + +# re: result attrs see top comment # pylint: disable=protected-access def _format_failure(failure): '''Return a list of pretty-formatted text entries describing a failure, including relevant information about it. Expect that the list of text entries will be joined @@ -101,11 +146,8 @@ def _format_failed_checks(checks): return stringc(pformat(checks), C.COLOR_ERROR) -# Reason: disable pylint protected-access because we need to access _* -# attributes of obj to implement this function. -# This is inspired by ansible.playbook.base.Base.dump_me. -# Status: permanently disabled unless Ansible's API changes. -# pylint: disable=protected-access +# This is inspired by ansible.playbook.base.Base.dump_me. +# re: play/task/block attrs see top comment # pylint: disable=protected-access def _get_play(obj): '''Given a task or block, recursively tries to find its parent play.''' if hasattr(obj, '_play'): diff --git a/roles/openshift_health_checker/library/aos_version.py b/roles/openshift_health_checker/library/aos_version.py index 191a4b107..4c205e48c 100755 --- a/roles/openshift_health_checker/library/aos_version.py +++ b/roles/openshift_health_checker/library/aos_version.py @@ -1,91 +1,203 @@ #!/usr/bin/python -# vim: expandtab:tabstop=4:shiftwidth=4 ''' -Ansible module for determining if multiple versions of an OpenShift package are -available, and if the version requested is available down to the given -precision. - -Multiple versions available suggest that multiple repos are enabled for the -different versions, which may cause installation problems. +Ansible module for yum-based systems determining if multiple releases +of an OpenShift package are available, and if the release requested +(if any) is available down to the given precision. + +For Enterprise, multiple releases available suggest that multiple repos +are enabled for the different releases, which may cause installation +problems. With Origin, however, this is a normal state of affairs as +all the releases are provided in a single repo with the expectation that +only the latest can be installed. + +Code in the openshift_version role contains a lot of logic to pin down +the exact package and image version to use and so does some validation +of release availability already. Without duplicating all that, we would +like the user to have a helpful error message if we detect things will +not work out right. Note that if openshift_release is not specified in +the inventory, the version comparison checks just pass. ''' -import yum # pylint: disable=import-error - from ansible.module_utils.basic import AnsibleModule +IMPORT_EXCEPTION = None +try: + import yum # pylint: disable=import-error +except ImportError as err: + IMPORT_EXCEPTION = err + + +class AosVersionException(Exception): + '''Base exception class for package version problems''' + def __init__(self, message, problem_pkgs=None): + Exception.__init__(self, message) + self.problem_pkgs = problem_pkgs + -def main(): # pylint: disable=missing-docstring,too-many-branches +def main(): + """Entrypoint for this Ansible module""" module = AnsibleModule( argument_spec=dict( - prefix=dict(required=True), # atomic-openshift, origin, ... - version=dict(required=True), + package_list=dict(type="list", required=True), ), supports_check_mode=True ) - def bail(error): # pylint: disable=missing-docstring - module.fail_json(msg=error) + if IMPORT_EXCEPTION: + module.fail_json(msg="aos_version module could not import yum: %s" % IMPORT_EXCEPTION) - rpm_prefix = module.params['prefix'] + # determine the packages we will look for + package_list = module.params['package_list'] + if not package_list: + module.fail_json(msg="package_list must not be empty") - if not rpm_prefix: - bail("prefix must not be empty") + # generate set with only the names of expected packages + expected_pkg_names = [p["name"] for p in package_list] + + # gather packages that require a multi_minor_release check + multi_minor_pkgs = [p for p in package_list if p["check_multi"]] + + # generate list of packages with a specified (non-empty) version + # should look like a version string with possibly many segments e.g. "3.4.1" + versioned_pkgs = [p for p in package_list if p["version"]] + + # get the list of packages available and complain if anything is wrong + try: + pkgs = _retrieve_available_packages(expected_pkg_names) + if versioned_pkgs: + _check_precise_version_found(pkgs, _to_dict(versioned_pkgs)) + _check_higher_version_found(pkgs, _to_dict(versioned_pkgs)) + if multi_minor_pkgs: + _check_multi_minor_release(pkgs, _to_dict(multi_minor_pkgs)) + except AosVersionException as excinfo: + module.fail_json(msg=str(excinfo)) + module.exit_json(changed=False) + +def _to_dict(pkg_list): + return {pkg["name"]: pkg for pkg in pkg_list} + + +def _retrieve_available_packages(expected_pkgs): + # search for package versions available for openshift pkgs yb = yum.YumBase() # pylint: disable=invalid-name - yb.conf.disable_excludes = ["all"] # assume the openshift excluder will be managed, ignore current state - - # search for package versions available for aos pkgs - expected_pkgs = [ - rpm_prefix, - rpm_prefix + '-master', - rpm_prefix + '-node', - ] + + # The openshift excluder prevents unintended updates to openshift + # packages by setting yum excludes on those packages. See: + # https://wiki.centos.org/SpecialInterestGroup/PaaS/OpenShift-Origin-Control-Updates + # Excludes are then disabled during an install or upgrade, but + # this check will most likely be running outside either. When we + # attempt to determine what packages are available via yum they may + # be excluded. So, for our purposes here, disable excludes to see + # what will really be available during an install or upgrade. + yb.conf.disable_excludes = ['all'] + try: pkgs = yb.pkgSack.returnPackages(patterns=expected_pkgs) - except yum.Errors.PackageSackError as e: # pylint: disable=invalid-name + except yum.Errors.PackageSackError as excinfo: # you only hit this if *none* of the packages are available - bail('Unable to find any OpenShift packages.\nCheck your subscription and repo settings.\n%s' % e) + raise AosVersionException('\n'.join([ + 'Unable to find any OpenShift packages.', + 'Check your subscription and repo settings.', + str(excinfo), + ])) + return pkgs + + +class PreciseVersionNotFound(AosVersionException): + """Exception for reporting packages not available at given version""" + def __init__(self, not_found): + msg = ['Not all of the required packages are available at their requested version'] + msg += ['{}:{} '.format(pkg["name"], pkg["version"]) for pkg in not_found] + msg += ['Please check your subscriptions and enabled repositories.'] + AosVersionException.__init__(self, '\n'.join(msg), not_found) + + +def _check_precise_version_found(pkgs, expected_pkgs_dict): + # see if any packages couldn't be found at requested release version + # we would like to verify that the latest available pkgs have however specific a version is given. + # so e.g. if there is a package version 3.4.1.5 the check passes; if only 3.4.0, it fails. - # determine what level of precision we're expecting for the version - expected_version = module.params['version'] - if expected_version.startswith('v'): # v3.3 => 3.3 - expected_version = expected_version[1:] - num_dots = expected_version.count('.') + pkgs_precise_version_found = set() + for pkg in pkgs: + if pkg.name not in expected_pkgs_dict: + continue + # does the version match, to the precision requested? + # and, is it strictly greater, at the precision requested? + expected_pkg_version = expected_pkgs_dict[pkg.name]["version"] + match_version = '.'.join(pkg.version.split('.')[:expected_pkg_version.count('.') + 1]) + if match_version == expected_pkg_version: + pkgs_precise_version_found.add(pkg.name) + + not_found = [] + for name, pkg in expected_pkgs_dict.items(): + if name not in pkgs_precise_version_found: + not_found.append(pkg) + + if not_found: + raise PreciseVersionNotFound(not_found) + + +class FoundHigherVersion(AosVersionException): + """Exception for reporting that a higher version than requested is available""" + def __init__(self, higher_found): + msg = ['Some required package(s) are available at a version', + 'that is higher than requested'] + msg += [' ' + name for name in higher_found] + msg += ['This will prevent installing the version you requested.'] + msg += ['Please check your enabled repositories or adjust openshift_release.'] + AosVersionException.__init__(self, '\n'.join(msg), higher_found) + + +def _check_higher_version_found(pkgs, expected_pkgs_dict): + expected_pkg_names = list(expected_pkgs_dict) + # see if any packages are available in a version higher than requested + higher_version_for_pkg = {} + for pkg in pkgs: + if pkg.name not in expected_pkg_names: + continue + expected_pkg_version = expected_pkgs_dict[pkg.name]["version"] + req_release_arr = [int(segment) for segment in expected_pkg_version.split(".")] + version = [int(segment) for segment in pkg.version.split(".")] + too_high = version[:len(req_release_arr)] > req_release_arr + higher_than_seen = version > higher_version_for_pkg.get(pkg.name, []) + if too_high and higher_than_seen: + higher_version_for_pkg[pkg.name] = version + + if higher_version_for_pkg: + higher_found = [] + for name, version in higher_version_for_pkg.items(): + higher_found.append(name + '-' + '.'.join(str(segment) for segment in version)) + raise FoundHigherVersion(higher_found) + + +class FoundMultiRelease(AosVersionException): + """Exception for reporting multiple minor releases found for same package""" + def __init__(self, multi_found): + msg = ['Multiple minor versions of these packages are available'] + msg += [' ' + name for name in multi_found] + msg += ["There should only be one OpenShift release repository enabled at a time."] + AosVersionException.__init__(self, '\n'.join(msg), multi_found) + + +def _check_multi_minor_release(pkgs, expected_pkgs_dict): + # see if any packages are available in more than one minor version pkgs_by_name_version = {} - pkgs_precise_version_found = {} for pkg in pkgs: - # get expected version precision - match_version = '.'.join(pkg.version.split('.')[:num_dots + 1]) - if match_version == expected_version: - pkgs_precise_version_found[pkg.name] = True - # get x.y version precision - minor_version = '.'.join(pkg.version.split('.')[:2]) + # keep track of x.y (minor release) versions seen + minor_release = '.'.join(pkg.version.split('.')[:2]) if pkg.name not in pkgs_by_name_version: - pkgs_by_name_version[pkg.name] = {} - pkgs_by_name_version[pkg.name][minor_version] = True + pkgs_by_name_version[pkg.name] = set() + pkgs_by_name_version[pkg.name].add(minor_release) - # see if any packages couldn't be found at requested version - # see if any packages are available in more than one minor version - not_found = [] multi_found = [] - for name in expected_pkgs: - if name not in pkgs_precise_version_found: - not_found.append(name) + for name in expected_pkgs_dict: if name in pkgs_by_name_version and len(pkgs_by_name_version[name]) > 1: multi_found.append(name) - if not_found: - msg = 'Not all of the required packages are available at requested version %s:\n' % expected_version - for name in not_found: - msg += ' %s\n' % name - bail(msg + 'Please check your subscriptions and enabled repositories.') - if multi_found: - msg = 'Multiple minor versions of these packages are available\n' - for name in multi_found: - msg += ' %s\n' % name - bail(msg + "There should only be one OpenShift version's repository enabled at a time.") - module.exit_json(changed=False) + if multi_found: + raise FoundMultiRelease(multi_found) if __name__ == '__main__': diff --git a/roles/openshift_health_checker/library/check_yum_update.py b/roles/openshift_health_checker/library/check_yum_update.py index 630ebc848..433795b67 100755 --- a/roles/openshift_health_checker/library/check_yum_update.py +++ b/roles/openshift_health_checker/library/check_yum_update.py @@ -1,5 +1,4 @@ #!/usr/bin/python -# vim: expandtab:tabstop=4:shiftwidth=4 ''' Ansible module to test whether a yum update or install will succeed, without actually performing it or running yum. diff --git a/roles/openshift_health_checker/library/etcdkeysize.py b/roles/openshift_health_checker/library/etcdkeysize.py new file mode 100644 index 000000000..620e82d87 --- /dev/null +++ b/roles/openshift_health_checker/library/etcdkeysize.py @@ -0,0 +1,122 @@ +#!/usr/bin/python +"""Ansible module that recursively determines if the size of a key in an etcd cluster exceeds a given limit.""" + +from ansible.module_utils.basic import AnsibleModule + + +try: + import etcd + + IMPORT_EXCEPTION_MSG = None +except ImportError as err: + IMPORT_EXCEPTION_MSG = str(err) + + from collections import namedtuple + EtcdMock = namedtuple("etcd", ["EtcdKeyNotFound"]) + etcd = EtcdMock(KeyError) + + +# pylint: disable=too-many-arguments +def check_etcd_key_size(client, key, size_limit, total_size=0, depth=0, depth_limit=1000, visited=None): + """Check size of an etcd path starting at given key. Returns tuple (string, bool)""" + if visited is None: + visited = set() + + if key in visited: + return 0, False + + visited.add(key) + + try: + result = client.read(key, recursive=False) + except etcd.EtcdKeyNotFound: + return 0, False + + size = 0 + limit_exceeded = False + + for node in result.leaves: + if depth >= depth_limit: + raise Exception("Maximum recursive stack depth ({}) exceeded.".format(depth_limit)) + + if size_limit and total_size + size > size_limit: + return size, True + + if not node.dir: + size += len(node.value) + continue + + key_size, limit_exceeded = check_etcd_key_size(client, node.key, + size_limit, + total_size + size, + depth + 1, + depth_limit, visited) + size += key_size + + max_limit_exceeded = limit_exceeded or (total_size + size > size_limit) + return size, max_limit_exceeded + + +def main(): # pylint: disable=missing-docstring,too-many-branches + module = AnsibleModule( + argument_spec=dict( + size_limit_bytes=dict(type="int", default=0), + paths=dict(type="list", default=["/openshift.io/images"]), + host=dict(type="str", default="127.0.0.1"), + port=dict(type="int", default=4001), + protocol=dict(type="str", default="http"), + version_prefix=dict(type="str", default=""), + allow_redirect=dict(type="bool", default=False), + cert=dict(type="dict", default=""), + ca_cert=dict(type="str", default=None), + ), + supports_check_mode=True + ) + + module.params["cert"] = ( + module.params["cert"]["cert"], + module.params["cert"]["key"], + ) + + size_limit = module.params.pop("size_limit_bytes") + paths = module.params.pop("paths") + + limit_exceeded = False + + try: + # pylint: disable=no-member + client = etcd.Client(**module.params) + except AttributeError as attrerr: + msg = str(attrerr) + if IMPORT_EXCEPTION_MSG: + msg = IMPORT_EXCEPTION_MSG + if "No module named etcd" in IMPORT_EXCEPTION_MSG: + # pylint: disable=redefined-variable-type + msg = ('Unable to import the python "etcd" dependency. ' + 'Make sure python-etcd is installed on the host.') + + module.exit_json( + failed=True, + changed=False, + size_limit_exceeded=limit_exceeded, + msg=msg, + ) + + return + + size = 0 + for path in paths: + path_size, limit_exceeded = check_etcd_key_size(client, path, size_limit - size) + size += path_size + + if limit_exceeded: + break + + module.exit_json( + changed=False, + size_limit_exceeded=limit_exceeded, + ) + + +if __name__ == '__main__': + main() diff --git a/roles/openshift_health_checker/library/ocutil.py b/roles/openshift_health_checker/library/ocutil.py new file mode 100644 index 000000000..2e60735d6 --- /dev/null +++ b/roles/openshift_health_checker/library/ocutil.py @@ -0,0 +1,74 @@ +#!/usr/bin/python +"""Interface to OpenShift oc command""" + +import os +import shlex +import shutil +import subprocess + +from ansible.module_utils.basic import AnsibleModule + + +ADDITIONAL_PATH_LOOKUPS = ['/usr/local/bin', os.path.expanduser('~/bin')] + + +def locate_oc_binary(): + """Find and return oc binary file""" + # https://github.com/openshift/openshift-ansible/issues/3410 + # oc can be in /usr/local/bin in some cases, but that may not + # be in $PATH due to ansible/sudo + paths = os.environ.get("PATH", os.defpath).split(os.pathsep) + ADDITIONAL_PATH_LOOKUPS + + oc_binary = 'oc' + + # Use shutil.which if it is available, otherwise fallback to a naive path search + try: + which_result = shutil.which(oc_binary, path=os.pathsep.join(paths)) + if which_result is not None: + oc_binary = which_result + except AttributeError: + for path in paths: + if os.path.exists(os.path.join(path, oc_binary)): + oc_binary = os.path.join(path, oc_binary) + break + + return oc_binary + + +def main(): + """Module that executes commands on a remote OpenShift cluster""" + + module = AnsibleModule( + argument_spec=dict( + namespace=dict(type="str", required=True), + config_file=dict(type="str", required=True), + cmd=dict(type="str", required=True), + extra_args=dict(type="list", default=[]), + ), + ) + + cmd = [ + locate_oc_binary(), + '--config', module.params["config_file"], + '-n', module.params["namespace"], + ] + shlex.split(module.params["cmd"]) + + failed = True + try: + cmd_result = subprocess.check_output(list(cmd), stderr=subprocess.STDOUT) + failed = False + except subprocess.CalledProcessError as exc: + cmd_result = '[rc {}] {}\n{}'.format(exc.returncode, ' '.join(exc.cmd), exc.output) + except OSError as exc: + # we get this when 'oc' is not there + cmd_result = str(exc) + + module.exit_json( + changed=False, + failed=failed, + result=cmd_result, + ) + + +if __name__ == '__main__': + main() diff --git a/roles/openshift_health_checker/library/rpm_version.py b/roles/openshift_health_checker/library/rpm_version.py new file mode 100644 index 000000000..8ea223055 --- /dev/null +++ b/roles/openshift_health_checker/library/rpm_version.py @@ -0,0 +1,127 @@ +#!/usr/bin/python +""" +Ansible module for rpm-based systems determining existing package version information in a host. +""" + +from ansible.module_utils.basic import AnsibleModule + +IMPORT_EXCEPTION = None +try: + import rpm # pylint: disable=import-error +except ImportError as err: + IMPORT_EXCEPTION = err # in tox test env, rpm import fails + + +class RpmVersionException(Exception): + """Base exception class for package version problems""" + def __init__(self, message, problem_pkgs=None): + Exception.__init__(self, message) + self.problem_pkgs = problem_pkgs + + +def main(): + """Entrypoint for this Ansible module""" + module = AnsibleModule( + argument_spec=dict( + package_list=dict(type="list", required=True), + ), + supports_check_mode=True + ) + + if IMPORT_EXCEPTION: + module.fail_json(msg="rpm_version module could not import rpm: %s" % IMPORT_EXCEPTION) + + # determine the packages we will look for + pkg_list = module.params['package_list'] + if not pkg_list: + module.fail_json(msg="package_list must not be empty") + + # get list of packages available and complain if any + # of them are missing or if any errors occur + try: + pkg_versions = _retrieve_expected_pkg_versions(_to_dict(pkg_list)) + _check_pkg_versions(pkg_versions, _to_dict(pkg_list)) + except RpmVersionException as excinfo: + module.fail_json(msg=str(excinfo)) + module.exit_json(changed=False) + + +def _to_dict(pkg_list): + return {pkg["name"]: pkg for pkg in pkg_list} + + +def _retrieve_expected_pkg_versions(expected_pkgs_dict): + """Search for installed packages matching given pkg names + and versions. Returns a dictionary: {pkg_name: [versions]}""" + + transaction = rpm.TransactionSet() + pkgs = {} + + for pkg_name in expected_pkgs_dict: + matched_pkgs = transaction.dbMatch("name", pkg_name) + if not matched_pkgs: + continue + + for header in matched_pkgs: + if header['name'] == pkg_name: + if pkg_name not in pkgs: + pkgs[pkg_name] = [] + + pkgs[pkg_name].append(header['version']) + + return pkgs + + +def _check_pkg_versions(found_pkgs_dict, expected_pkgs_dict): + invalid_pkg_versions = {} + not_found_pkgs = [] + + for pkg_name, pkg in expected_pkgs_dict.items(): + if not found_pkgs_dict.get(pkg_name): + not_found_pkgs.append(pkg_name) + continue + + found_versions = [_parse_version(version) for version in found_pkgs_dict[pkg_name]] + expected_version = _parse_version(pkg["version"]) + if expected_version not in found_versions: + invalid_pkg_versions[pkg_name] = { + "found_versions": found_versions, + "required_version": expected_version, + } + + if not_found_pkgs: + raise RpmVersionException( + '\n'.join([ + "The following packages were not found to be installed: {}".format('\n '.join([ + "{}".format(pkg) + for pkg in not_found_pkgs + ])) + ]), + not_found_pkgs, + ) + + if invalid_pkg_versions: + raise RpmVersionException( + '\n '.join([ + "The following packages were found to be installed with an incorrect version: {}".format('\n'.join([ + " \n{}\n Required version: {}\n Found versions: {}".format( + pkg_name, + pkg["required_version"], + ', '.join([version for version in pkg["found_versions"]])) + for pkg_name, pkg in invalid_pkg_versions.items() + ])) + ]), + invalid_pkg_versions, + ) + + +def _parse_version(version_str): + segs = version_str.split('.') + if not segs or len(segs) <= 2: + return version_str + + return '.'.join(segs[0:2]) + + +if __name__ == '__main__': + main() diff --git a/roles/openshift_health_checker/meta/main.yml b/roles/openshift_health_checker/meta/main.yml index 0bbeadd34..4d141974c 100644 --- a/roles/openshift_health_checker/meta/main.yml +++ b/roles/openshift_health_checker/meta/main.yml @@ -1,3 +1,5 @@ --- dependencies: - role: openshift_facts + - role: openshift_repos + - role: openshift_version diff --git a/roles/openshift_health_checker/openshift_checks/__init__.py b/roles/openshift_health_checker/openshift_checks/__init__.py index be63d864a..5c9949ced 100644 --- a/roles/openshift_health_checker/openshift_checks/__init__.py +++ b/roles/openshift_health_checker/openshift_checks/__init__.py @@ -66,16 +66,26 @@ class OpenShiftCheck(object): LOADER_EXCLUDES = ( "__init__.py", "mixins.py", + "logging.py", ) -def load_checks(): +def load_checks(path=None, subpkg=""): """Dynamically import all check modules for the side effect of registering checks.""" - return [ - import_module(__package__ + "." + name[:-3]) - for name in os.listdir(os.path.dirname(__file__)) - if name.endswith(".py") and name not in LOADER_EXCLUDES - ] + if path is None: + path = os.path.dirname(__file__) + + modules = [] + + for name in os.listdir(path): + if os.path.isdir(os.path.join(path, name)): + modules = modules + load_checks(os.path.join(path, name), subpkg + "." + name) + continue + + if name.endswith(".py") and name not in LOADER_EXCLUDES: + modules.append(import_module(__package__ + subpkg + "." + name[:-3])) + + return modules def get_var(task_vars, *keys, **kwargs): diff --git a/roles/openshift_health_checker/openshift_checks/disk_availability.py b/roles/openshift_health_checker/openshift_checks/disk_availability.py new file mode 100644 index 000000000..e93e81efa --- /dev/null +++ b/roles/openshift_health_checker/openshift_checks/disk_availability.py @@ -0,0 +1,115 @@ +"""Check that there is enough disk space in predefined paths.""" + +import os.path +import tempfile + +from openshift_checks import OpenShiftCheck, OpenShiftCheckException, get_var + + +class DiskAvailability(OpenShiftCheck): + """Check that recommended disk space is available before a first-time install.""" + + name = "disk_availability" + tags = ["preflight"] + + # Values taken from the official installation documentation: + # https://docs.openshift.org/latest/install_config/install/prerequisites.html#system-requirements + recommended_disk_space_bytes = { + '/var': { + 'masters': 40 * 10**9, + 'nodes': 15 * 10**9, + 'etcd': 20 * 10**9, + }, + # Used to copy client binaries into, + # see roles/openshift_cli/library/openshift_container_binary_sync.py. + '/usr/local/bin': { + 'masters': 1 * 10**9, + 'nodes': 1 * 10**9, + 'etcd': 1 * 10**9, + }, + # Used as temporary storage in several cases. + tempfile.gettempdir(): { + 'masters': 1 * 10**9, + 'nodes': 1 * 10**9, + 'etcd': 1 * 10**9, + }, + } + + @classmethod + def is_active(cls, task_vars): + """Skip hosts that do not have recommended disk space requirements.""" + group_names = get_var(task_vars, "group_names", default=[]) + active_groups = set() + for recommendation in cls.recommended_disk_space_bytes.values(): + active_groups.update(recommendation.keys()) + has_disk_space_recommendation = bool(active_groups.intersection(group_names)) + return super(DiskAvailability, cls).is_active(task_vars) and has_disk_space_recommendation + + def run(self, tmp, task_vars): + group_names = get_var(task_vars, "group_names") + ansible_mounts = get_var(task_vars, "ansible_mounts") + ansible_mounts = {mount['mount']: mount for mount in ansible_mounts} + + user_config = get_var(task_vars, "openshift_check_min_host_disk_gb", default={}) + try: + # For backwards-compatibility, if openshift_check_min_host_disk_gb + # is a number, then it overrides the required config for '/var'. + number = float(user_config) + user_config = { + '/var': { + 'masters': number, + 'nodes': number, + 'etcd': number, + }, + } + except TypeError: + # If it is not a number, then it should be a nested dict. + pass + + # TODO: as suggested in + # https://github.com/openshift/openshift-ansible/pull/4436#discussion_r122180021, + # maybe we could support checking disk availability in paths that are + # not part of the official recommendation but present in the user + # configuration. + for path, recommendation in self.recommended_disk_space_bytes.items(): + free_bytes = self.free_bytes(path, ansible_mounts) + recommended_bytes = max(recommendation.get(name, 0) for name in group_names) + + config = user_config.get(path, {}) + # NOTE: the user config is in GB, but we compare bytes, thus the + # conversion. + config_bytes = max(config.get(name, 0) for name in group_names) * 10**9 + recommended_bytes = config_bytes or recommended_bytes + + if free_bytes < recommended_bytes: + free_gb = float(free_bytes) / 10**9 + recommended_gb = float(recommended_bytes) / 10**9 + return { + 'failed': True, + 'msg': ( + 'Available disk space in "{}" ({:.1f} GB) ' + 'is below minimum recommended ({:.1f} GB)' + ).format(path, free_gb, recommended_gb) + } + + return {} + + @staticmethod + def free_bytes(path, ansible_mounts): + """Return the size available in path based on ansible_mounts.""" + mount_point = path + # arbitry value to prevent an infinite loop, in the unlike case that '/' + # is not in ansible_mounts. + max_depth = 32 + while mount_point not in ansible_mounts and max_depth > 0: + mount_point = os.path.dirname(mount_point) + max_depth -= 1 + + try: + free_bytes = ansible_mounts[mount_point]['size_available'] + except KeyError: + known_mounts = ', '.join('"{}"'.format(mount) for mount in sorted(ansible_mounts)) or 'none' + msg = 'Unable to determine disk availability for "{}". Known mount points: {}.' + raise OpenShiftCheckException(msg.format(path, known_mounts)) + + return free_bytes diff --git a/roles/openshift_health_checker/openshift_checks/docker_image_availability.py b/roles/openshift_health_checker/openshift_checks/docker_image_availability.py index cce289b95..26bf4c09b 100644 --- a/roles/openshift_health_checker/openshift_checks/docker_image_availability.py +++ b/roles/openshift_health_checker/openshift_checks/docker_image_availability.py @@ -1,8 +1,25 @@ -# pylint: disable=missing-docstring +"""Check that required Docker images are available.""" + from openshift_checks import OpenShiftCheck, get_var +from openshift_checks.mixins import DockerHostMixin + + +NODE_IMAGE_SUFFIXES = ["haproxy-router", "docker-registry", "deployer", "pod"] +DEPLOYMENT_IMAGE_INFO = { + "origin": { + "namespace": "openshift", + "name": "origin", + "registry_console_image": "cockpit/kubernetes", + }, + "openshift-enterprise": { + "namespace": "openshift3", + "name": "ose", + "registry_console_image": "registry.access.redhat.com/openshift3/registry-console", + }, +} -class DockerImageAvailability(OpenShiftCheck): +class DockerImageAvailability(DockerHostMixin, OpenShiftCheck): """Check that required Docker images are available. This check attempts to ensure that required docker images are @@ -12,42 +29,36 @@ class DockerImageAvailability(OpenShiftCheck): name = "docker_image_availability" tags = ["preflight"] + dependencies = ["skopeo", "python-docker-py"] - skopeo_image = "openshift/openshift-ansible" - - # FIXME(juanvallejo): we should consider other possible values of - # `deployment_type` (the key here). See - # https://github.com/openshift/openshift-ansible/blob/8e26f8c/roles/openshift_repos/vars/main.yml#L7 - docker_image_base = { - "origin": { - "repo": "openshift", - "image": "origin", - }, - "openshift-enterprise": { - "repo": "openshift3", - "image": "ose", - }, - } + @classmethod + def is_active(cls, task_vars): + """Skip hosts with unsupported deployment types.""" + deployment_type = get_var(task_vars, "openshift_deployment_type") + has_valid_deployment_type = deployment_type in DEPLOYMENT_IMAGE_INFO - def run(self, tmp, task_vars): - required_images = self.required_images(task_vars) - missing_images = set(required_images) - set(self.local_images(required_images, task_vars)) + return super(DockerImageAvailability, cls).is_active(task_vars) and has_valid_deployment_type - # exit early if all images were found locally - if not missing_images: - return {"changed": False} - - msg, failed, changed = self.update_skopeo_image(task_vars) - - # exit early if Skopeo update fails + def run(self, tmp, task_vars): + msg, failed, changed = self.ensure_dependencies(task_vars) if failed: return { "failed": True, "changed": changed, - "msg": "Failed to update Skopeo image ({img_name}). {msg}".format(img_name=self.skopeo_image, msg=msg), + "msg": "Some dependencies are required in order to check Docker image availability.\n" + msg } + required_images = self.required_images(task_vars) + missing_images = set(required_images) - set(self.local_images(required_images, task_vars)) + + # exit early if all images were found locally + if not missing_images: + return {"changed": changed} + registries = self.known_docker_registries(task_vars) + if not registries: + return {"failed": True, "msg": "Unable to retrieve any docker registries.", "changed": changed} + available_images = self.available_images(missing_images, registries, task_vars) unavailable_images = set(missing_images) - set(available_images) @@ -55,43 +66,63 @@ class DockerImageAvailability(OpenShiftCheck): return { "failed": True, "msg": ( - "One or more required images are not available: {}.\n" + "One or more required Docker images are not available:\n {}\n" "Configured registries: {}" - ).format(", ".join(sorted(unavailable_images)), ", ".join(registries)), + ).format(",\n ".join(sorted(unavailable_images)), ", ".join(registries)), "changed": changed, } return {"changed": changed} - def required_images(self, task_vars): - deployment_type = get_var(task_vars, "deployment_type") - # FIXME(juanvallejo): we should handle gracefully with a proper error - # message when given an unexpected value for `deployment_type`. - image_base_name = self.docker_image_base[deployment_type] - - openshift_release = get_var(task_vars, "openshift_release") - # FIXME(juanvallejo): this variable is not required when the - # installation is non-containerized. The example inventories have it - # commented out. We should handle gracefully and with a proper error - # message when this variable is required and not set. - openshift_image_tag = get_var(task_vars, "openshift_image_tag") - - is_containerized = get_var(task_vars, "openshift", "common", "is_containerized") - - if is_containerized: - images = set(self.containerized_docker_images(image_base_name, openshift_release)) - else: - images = set(self.rpm_docker_images(image_base_name, openshift_release)) - - # append images with qualified image tags to our list of required images. - # these are images with a (v0.0.0.0) tag, rather than a standard release - # format tag (v0.0). We want to check this set in both containerized and - # non-containerized installations. - images.update( - self.qualified_docker_images(self.image_from_base_name(image_base_name), "v" + openshift_image_tag) - ) - - return images + @staticmethod + def required_images(task_vars): + """ + Determine which images we expect to need for this host. + Returns: a set of required images like 'openshift/origin:v3.6' + + The thorny issue of determining the image names from the variables is under consideration + via https://github.com/openshift/openshift-ansible/issues/4415 + + For now we operate as follows: + * For containerized components (master, node, ...) we look at the deployment type and + use openshift/origin or openshift3/ose as the base for those component images. The + version is openshift_image_tag as determined by the openshift_version role. + * For OpenShift-managed infrastructure (router, registry...) we use oreg_url if + it is defined; otherwise we again use the base that depends on the deployment type. + Registry is not included in constructed images. It may be in oreg_url or etcd image. + """ + required = set() + deployment_type = get_var(task_vars, "openshift_deployment_type") + host_groups = get_var(task_vars, "group_names") + image_tag = get_var(task_vars, "openshift_image_tag") + image_info = DEPLOYMENT_IMAGE_INFO[deployment_type] + if not image_info: + return required + + # template for images that run on top of OpenShift + image_url = "{}/{}-{}:{}".format(image_info["namespace"], image_info["name"], "${component}", "${version}") + image_url = get_var(task_vars, "oreg_url", default="") or image_url + if 'nodes' in host_groups: + for suffix in NODE_IMAGE_SUFFIXES: + required.add(image_url.replace("${component}", suffix).replace("${version}", image_tag)) + # The registry-console is for some reason not prefixed with ose- like the other components. + # Nor is it versioned the same, so just look for latest. + # Also a completely different name is used for Origin. + required.add(image_info["registry_console_image"]) + + # images for containerized components + if get_var(task_vars, "openshift", "common", "is_containerized"): + components = set() + if 'nodes' in host_groups: + components.update(["node", "openvswitch"]) + if 'masters' in host_groups: # name is "origin" or "ose" + components.add(image_info["name"]) + for component in components: + required.add("{}/{}:{}".format(image_info["namespace"], component, image_tag)) + if 'etcd' in host_groups: # special case, note it is the same for origin/enterprise + required.add("registry.access.redhat.com/rhel7/etcd") # and no image tag + + return required def local_images(self, images, task_vars): """Filter a list of images and return those available locally.""" @@ -101,79 +132,46 @@ class DockerImageAvailability(OpenShiftCheck): ] def is_image_local(self, image, task_vars): - result = self.module_executor("docker_image_facts", {"name": image}, task_vars) + """Check if image is already in local docker index.""" + result = self.execute_module("docker_image_facts", {"name": image}, task_vars=task_vars) if result.get("failed", False): return False return bool(result.get("images", [])) - def known_docker_registries(self, task_vars): - result = self.module_executor("docker_info", {}, task_vars) + @staticmethod + def known_docker_registries(task_vars): + """Build a list of docker registries available according to inventory vars.""" + docker_facts = get_var(task_vars, "openshift", "docker") + regs = set(docker_facts["additional_registries"]) - if result.get("failed", False): - return [] + deployment_type = get_var(task_vars, "openshift_deployment_type") + if deployment_type == "origin": + regs.update(["docker.io"]) + elif "enterprise" in deployment_type: + regs.update(["registry.access.redhat.com"]) - # FIXME(juanvallejo): wrong default type, result["info"] is expected to - # contain a dictionary (see how we call `docker_info.get` below). - docker_info = result.get("info", "") - return [registry.get("Name", "") for registry in docker_info.get("Registries", {})] + return list(regs) def available_images(self, images, registries, task_vars): """Inspect existing images using Skopeo and return all images successfully inspected.""" return [ image for image in images - if self.is_image_available(image, registries, task_vars) + if self.is_available_skopeo_image(image, registries, task_vars) ] - def is_image_available(self, image, registries, task_vars): + def is_available_skopeo_image(self, image, registries, task_vars): + """Use Skopeo to determine if required image exists in known registry(s).""" + + # if image does already includes a registry, just use that + if image.count("/") > 1: + registry, image = image.split("/", 1) + registries = [registry] + for registry in registries: - if self.is_available_skopeo_image(image, registry, task_vars): + args = {"_raw_params": "skopeo inspect --tls-verify=false docker://{}/{}".format(registry, image)} + result = self.execute_module("command", args, task_vars=task_vars) + if result.get("rc", 0) == 0 and not result.get("failed"): return True return False - - def is_available_skopeo_image(self, image, registry, task_vars): - """Uses Skopeo to determine if required image exists in a given registry.""" - - cmd_str = "skopeo inspect docker://{registry}/{image}".format( - registry=registry, - image=image, - ) - - args = { - "name": "skopeo_inspect", - "image": self.skopeo_image, - "command": cmd_str, - "detach": False, - "cleanup": True, - } - result = self.module_executor("docker_container", args, task_vars) - return result.get("failed", False) - - def containerized_docker_images(self, base_name, version): - return [ - "{image}:{version}".format(image=self.image_from_base_name(base_name), version=version) - ] - - @staticmethod - def rpm_docker_images(base, version): - return [ - "{image_repo}/registry-console:{version}".format(image_repo=base["repo"], version=version) - ] - - @staticmethod - def qualified_docker_images(image_name, version): - return [ - "{}-{}:{}".format(image_name, component, version) - for component in "haproxy-router docker-registry deployer pod".split() - ] - - @staticmethod - def image_from_base_name(base): - return "".join([base["repo"], "/", base["image"]]) - - # ensures that the skopeo docker image exists, and updates it - # with latest if image was already present locally. - def update_skopeo_image(self, task_vars): - result = self.module_executor("docker_image", {"name": self.skopeo_image}, task_vars) - return result.get("msg", ""), result.get("failed", False), result.get("changed", False) diff --git a/roles/openshift_health_checker/openshift_checks/docker_storage.py b/roles/openshift_health_checker/openshift_checks/docker_storage.py new file mode 100644 index 000000000..8d0fbcc9c --- /dev/null +++ b/roles/openshift_health_checker/openshift_checks/docker_storage.py @@ -0,0 +1,185 @@ +"""Check Docker storage driver and usage.""" +import json +import re +from openshift_checks import OpenShiftCheck, OpenShiftCheckException, get_var +from openshift_checks.mixins import DockerHostMixin + + +class DockerStorage(DockerHostMixin, OpenShiftCheck): + """Check Docker storage driver compatibility. + + This check ensures that Docker is using a supported storage driver, + and that loopback is not being used (if using devicemapper). + Also that storage usage is not above threshold. + """ + + name = "docker_storage" + tags = ["pre-install", "health", "preflight"] + + dependencies = ["python-docker-py"] + storage_drivers = ["devicemapper", "overlay2"] + max_thinpool_data_usage_percent = 90.0 + max_thinpool_meta_usage_percent = 90.0 + + # pylint: disable=too-many-return-statements + # Reason: permanent stylistic exception; + # it is clearer to return on failures and there are just many ways to fail here. + def run(self, tmp, task_vars): + msg, failed, changed = self.ensure_dependencies(task_vars) + if failed: + return { + "failed": True, + "changed": changed, + "msg": "Some dependencies are required in order to query docker storage on host:\n" + msg + } + + # attempt to get the docker info hash from the API + info = self.execute_module("docker_info", {}, task_vars=task_vars) + if info.get("failed"): + return {"failed": True, "changed": changed, + "msg": "Failed to query Docker API. Is docker running on this host?"} + if not info.get("info"): # this would be very strange + return {"failed": True, "changed": changed, + "msg": "Docker API query missing info:\n{}".format(json.dumps(info))} + info = info["info"] + + # check if the storage driver we saw is valid + driver = info.get("Driver", "[NONE]") + if driver not in self.storage_drivers: + msg = ( + "Detected unsupported Docker storage driver '{driver}'.\n" + "Supported storage drivers are: {drivers}" + ).format(driver=driver, drivers=', '.join(self.storage_drivers)) + return {"failed": True, "changed": changed, "msg": msg} + + # driver status info is a list of tuples; convert to dict and validate based on driver + driver_status = {item[0]: item[1] for item in info.get("DriverStatus", [])} + if driver == "devicemapper": + if driver_status.get("Data loop file"): + msg = ( + "Use of loopback devices with the Docker devicemapper storage driver\n" + "(the default storage configuration) is unsupported in production.\n" + "Please use docker-storage-setup to configure a backing storage volume.\n" + "See http://red.ht/2rNperO for further information." + ) + return {"failed": True, "changed": changed, "msg": msg} + result = self._check_dm_usage(driver_status, task_vars) + result['changed'] = result.get('changed', False) or changed + return result + + # TODO(lmeyer): determine how to check usage for overlay2 + + return {"changed": changed} + + def _check_dm_usage(self, driver_status, task_vars): + """ + Backing assumptions: We expect devicemapper to be backed by an auto-expanding thin pool + implemented as an LV in an LVM2 VG. This is how docker-storage-setup currently configures + devicemapper storage. The LV is "thin" because it does not use all available storage + from its VG, instead expanding as needed; so to determine available space, we gather + current usage as the Docker API reports for the driver as well as space available for + expansion in the pool's VG. + Usage within the LV is divided into pools allocated to data and metadata, either of which + could run out of space first; so we check both. + """ + vals = dict( + vg_free=self._get_vg_free(driver_status.get("Pool Name"), task_vars), + data_used=driver_status.get("Data Space Used"), + data_total=driver_status.get("Data Space Total"), + metadata_used=driver_status.get("Metadata Space Used"), + metadata_total=driver_status.get("Metadata Space Total"), + ) + + # convert all human-readable strings to bytes + for key, value in vals.copy().items(): + try: + vals[key + "_bytes"] = self._convert_to_bytes(value) + except ValueError as err: # unlikely to hit this from API info, but just to be safe + return { + "failed": True, + "values": vals, + "msg": "Could not interpret {} value '{}' as bytes: {}".format(key, value, str(err)) + } + + # determine the threshold percentages which usage should not exceed + for name, default in [("data", self.max_thinpool_data_usage_percent), + ("metadata", self.max_thinpool_meta_usage_percent)]: + percent = get_var(task_vars, "max_thinpool_" + name + "_usage_percent", default=default) + try: + vals[name + "_threshold"] = float(percent) + except ValueError: + return { + "failed": True, + "msg": "Specified thinpool {} usage limit '{}' is not a percentage".format(name, percent) + } + + # test whether the thresholds are exceeded + messages = [] + for name in ["data", "metadata"]: + vals[name + "_pct_used"] = 100 * vals[name + "_used_bytes"] / ( + vals[name + "_total_bytes"] + vals["vg_free_bytes"]) + if vals[name + "_pct_used"] > vals[name + "_threshold"]: + messages.append( + "Docker thinpool {name} usage percentage {pct:.1f} " + "is higher than threshold {thresh:.1f}.".format( + name=name, + pct=vals[name + "_pct_used"], + thresh=vals[name + "_threshold"], + )) + vals["failed"] = True + + vals["msg"] = "\n".join(messages or ["Thinpool usage is within thresholds."]) + return vals + + def _get_vg_free(self, pool, task_vars): + # Determine which VG to examine according to the pool name, the only indicator currently + # available from the Docker API driver info. We assume a name that looks like + # "vg--name-docker--pool"; vg and lv names with inner hyphens doubled, joined by a hyphen. + match = re.match(r'((?:[^-]|--)+)-(?!-)', pool) # matches up to the first single hyphen + if not match: # unlikely, but... be clear if we assumed wrong + raise OpenShiftCheckException( + "This host's Docker reports it is using a storage pool named '{}'.\n" + "However this name does not have the expected format of 'vgname-lvname'\n" + "so the available storage in the VG cannot be determined.".format(pool) + ) + vg_name = match.groups()[0].replace("--", "-") + vgs_cmd = "/sbin/vgs --noheadings -o vg_free --units g --select vg_name=" + vg_name + # should return free space like " 12.00g" if the VG exists; empty if it does not + + ret = self.execute_module("command", {"_raw_params": vgs_cmd}, task_vars=task_vars) + if ret.get("failed") or ret.get("rc", 0) != 0: + raise OpenShiftCheckException( + "Is LVM installed? Failed to run /sbin/vgs " + "to determine docker storage usage:\n" + ret.get("msg", "") + ) + size = ret.get("stdout", "").strip() + if not size: + raise OpenShiftCheckException( + "This host's Docker reports it is using a storage pool named '{pool}'.\n" + "which we expect to come from local VG '{vg}'.\n" + "However, /sbin/vgs did not find this VG. Is Docker for this host" + "running and using the storage on the host?".format(pool=pool, vg=vg_name) + ) + return size + + @staticmethod + def _convert_to_bytes(string): + units = dict( + b=1, + k=1024, + m=1024**2, + g=1024**3, + t=1024**4, + p=1024**5, + ) + string = string or "" + match = re.match(r'(\d+(?:\.\d+)?)\s*(\w)?', string) # float followed by optional unit + if not match: + raise ValueError("Cannot convert to a byte size: " + string) + + number, unit = match.groups() + multiplier = 1 if not unit else units.get(unit.lower()) + if not multiplier: + raise ValueError("Cannot convert to a byte size: " + string) + + return float(number) * multiplier diff --git a/roles/openshift_health_checker/openshift_checks/etcd_imagedata_size.py b/roles/openshift_health_checker/openshift_checks/etcd_imagedata_size.py new file mode 100644 index 000000000..c04a69765 --- /dev/null +++ b/roles/openshift_health_checker/openshift_checks/etcd_imagedata_size.py @@ -0,0 +1,84 @@ +""" +Ansible module for determining if the size of OpenShift image data exceeds a specified limit in an etcd cluster. +""" + +from openshift_checks import OpenShiftCheck, OpenShiftCheckException, get_var + + +class EtcdImageDataSize(OpenShiftCheck): + """Check that total size of OpenShift image data does not exceed the recommended limit in an etcd cluster""" + + name = "etcd_imagedata_size" + tags = ["etcd"] + + def run(self, tmp, task_vars): + etcd_mountpath = self._get_etcd_mountpath(get_var(task_vars, "ansible_mounts")) + etcd_avail_diskspace = etcd_mountpath["size_available"] + etcd_total_diskspace = etcd_mountpath["size_total"] + + etcd_imagedata_size_limit = get_var(task_vars, + "etcd_max_image_data_size_bytes", + default=int(0.5 * float(etcd_total_diskspace - etcd_avail_diskspace))) + + etcd_is_ssl = get_var(task_vars, "openshift", "master", "etcd_use_ssl", default=False) + etcd_port = get_var(task_vars, "openshift", "master", "etcd_port", default=2379) + etcd_hosts = get_var(task_vars, "openshift", "master", "etcd_hosts") + + config_base = get_var(task_vars, "openshift", "common", "config_base") + + cert = task_vars.get("etcd_client_cert", config_base + "/master/master.etcd-client.crt") + key = task_vars.get("etcd_client_key", config_base + "/master/master.etcd-client.key") + ca_cert = task_vars.get("etcd_client_ca_cert", config_base + "/master/master.etcd-ca.crt") + + for etcd_host in list(etcd_hosts): + args = { + "size_limit_bytes": etcd_imagedata_size_limit, + "paths": ["/openshift.io/images", "/openshift.io/imagestreams"], + "host": etcd_host, + "port": etcd_port, + "protocol": "https" if etcd_is_ssl else "http", + "version_prefix": "/v2", + "allow_redirect": True, + "ca_cert": ca_cert, + "cert": { + "cert": cert, + "key": key, + }, + } + + etcdkeysize = self.module_executor("etcdkeysize", args, task_vars) + + if etcdkeysize.get("rc", 0) != 0 or etcdkeysize.get("failed"): + msg = 'Failed to retrieve stats for etcd host "{host}": {reason}' + reason = etcdkeysize.get("msg") + if etcdkeysize.get("module_stderr"): + reason = etcdkeysize["module_stderr"] + + msg = msg.format(host=etcd_host, reason=reason) + return {"failed": True, "changed": False, "msg": msg} + + if etcdkeysize["size_limit_exceeded"]: + limit = self._to_gigabytes(etcd_imagedata_size_limit) + msg = ("The size of OpenShift image data stored in etcd host " + "\"{host}\" exceeds the maximum recommended limit of {limit:.2f} GB. " + "Use the `oadm prune images` command to cleanup unused Docker images.") + return {"failed": True, "msg": msg.format(host=etcd_host, limit=limit)} + + return {"changed": False} + + @staticmethod + def _get_etcd_mountpath(ansible_mounts): + valid_etcd_mount_paths = ["/var/lib/etcd", "/var/lib", "/var", "/"] + + mount_for_path = {mnt.get("mount"): mnt for mnt in ansible_mounts} + for path in valid_etcd_mount_paths: + if path in mount_for_path: + return mount_for_path[path] + + paths = ', '.join(sorted(mount_for_path)) or 'none' + msg = "Unable to determine a valid etcd mountpath. Paths mounted: {}.".format(paths) + raise OpenShiftCheckException(msg) + + @staticmethod + def _to_gigabytes(byte_size): + return float(byte_size) / 10.0**9 diff --git a/roles/openshift_health_checker/openshift_checks/etcd_volume.py b/roles/openshift_health_checker/openshift_checks/etcd_volume.py new file mode 100644 index 000000000..7452c9cc1 --- /dev/null +++ b/roles/openshift_health_checker/openshift_checks/etcd_volume.py @@ -0,0 +1,58 @@ +"""A health check for OpenShift clusters.""" + +from openshift_checks import OpenShiftCheck, OpenShiftCheckException, get_var + + +class EtcdVolume(OpenShiftCheck): + """Ensures etcd storage usage does not exceed a given threshold.""" + + name = "etcd_volume" + tags = ["etcd", "health"] + + # Default device usage threshold. Value should be in the range [0, 100]. + default_threshold_percent = 90 + # Where to find ectd data, higher priority first. + supported_mount_paths = ["/var/lib/etcd", "/var/lib", "/var", "/"] + + @classmethod + def is_active(cls, task_vars): + etcd_hosts = get_var(task_vars, "groups", "etcd", default=[]) or get_var(task_vars, "groups", "masters", + default=[]) or [] + is_etcd_host = get_var(task_vars, "ansible_ssh_host") in etcd_hosts + return super(EtcdVolume, cls).is_active(task_vars) and is_etcd_host + + def run(self, tmp, task_vars): + mount_info = self._etcd_mount_info(task_vars) + available = mount_info["size_available"] + total = mount_info["size_total"] + used = total - available + + threshold = get_var( + task_vars, + "etcd_device_usage_threshold_percent", + default=self.default_threshold_percent + ) + + used_percent = 100.0 * used / total + + if used_percent > threshold: + device = mount_info.get("device", "unknown") + mount = mount_info.get("mount", "unknown") + msg = "etcd storage usage ({:.1f}%) is above threshold ({:.1f}%). Device: {}, mount: {}.".format( + used_percent, threshold, device, mount + ) + return {"failed": True, "msg": msg} + + return {"changed": False} + + def _etcd_mount_info(self, task_vars): + ansible_mounts = get_var(task_vars, "ansible_mounts") + mounts = {mnt.get("mount"): mnt for mnt in ansible_mounts} + + for path in self.supported_mount_paths: + if path in mounts: + return mounts[path] + + paths = ', '.join(sorted(mounts)) or 'none' + msg = "Unable to find etcd storage mount point. Paths mounted: {}.".format(paths) + raise OpenShiftCheckException(msg) diff --git a/roles/openshift_health_checker/openshift_checks/logging/__init__.py b/roles/openshift_health_checker/openshift_checks/logging/__init__.py new file mode 100644 index 000000000..e69de29bb --- /dev/null +++ b/roles/openshift_health_checker/openshift_checks/logging/__init__.py diff --git a/roles/openshift_health_checker/openshift_checks/logging/curator.py b/roles/openshift_health_checker/openshift_checks/logging/curator.py new file mode 100644 index 000000000..c9fc59896 --- /dev/null +++ b/roles/openshift_health_checker/openshift_checks/logging/curator.py @@ -0,0 +1,61 @@ +""" +Module for performing checks on an Curator logging deployment +""" + +from openshift_checks import get_var +from openshift_checks.logging.logging import LoggingCheck + + +class Curator(LoggingCheck): + """Module that checks an integrated logging Curator deployment""" + + name = "curator" + tags = ["health", "logging"] + + logging_namespace = None + + def run(self, tmp, task_vars): + """Check various things and gather errors. Returns: result as hash""" + + self.logging_namespace = get_var(task_vars, "openshift_logging_namespace", default="logging") + curator_pods, error = super(Curator, self).get_pods_for_component( + self.module_executor, + self.logging_namespace, + "curator", + task_vars + ) + if error: + return {"failed": True, "changed": False, "msg": error} + check_error = self.check_curator(curator_pods) + + if check_error: + msg = ("The following Curator deployment issue was found:" + "\n-------\n" + "{}".format(check_error)) + return {"failed": True, "changed": False, "msg": msg} + + # TODO(lmeyer): run it all again for the ops cluster + return {"failed": False, "changed": False, "msg": 'No problems found with Curator deployment.'} + + def check_curator(self, pods): + """Check to see if curator is up and working. Returns: error string""" + if not pods: + return ( + "There are no Curator pods for the logging stack,\n" + "so nothing will prune Elasticsearch indexes.\n" + "Is Curator correctly deployed?" + ) + + not_running = super(Curator, self).not_running_pods(pods) + if len(not_running) == len(pods): + return ( + "The Curator pod is not currently in a running state,\n" + "so Elasticsearch indexes may increase without bound." + ) + if len(pods) - len(not_running) > 1: + return ( + "There is more than one Curator pod running. This should not normally happen.\n" + "Although this doesn't cause any problems, you may want to investigate." + ) + + return None diff --git a/roles/openshift_health_checker/openshift_checks/logging/elasticsearch.py b/roles/openshift_health_checker/openshift_checks/logging/elasticsearch.py new file mode 100644 index 000000000..01cb35b81 --- /dev/null +++ b/roles/openshift_health_checker/openshift_checks/logging/elasticsearch.py @@ -0,0 +1,217 @@ +""" +Module for performing checks on an Elasticsearch logging deployment +""" + +import json +import re + +from openshift_checks import get_var +from openshift_checks.logging.logging import LoggingCheck + + +class Elasticsearch(LoggingCheck): + """Module that checks an integrated logging Elasticsearch deployment""" + + name = "elasticsearch" + tags = ["health", "logging"] + + logging_namespace = None + + def run(self, tmp, task_vars): + """Check various things and gather errors. Returns: result as hash""" + + self.logging_namespace = get_var(task_vars, "openshift_logging_namespace", default="logging") + es_pods, error = super(Elasticsearch, self).get_pods_for_component( + self.execute_module, + self.logging_namespace, + "es", + task_vars, + ) + if error: + return {"failed": True, "changed": False, "msg": error} + check_error = self.check_elasticsearch(es_pods, task_vars) + + if check_error: + msg = ("The following Elasticsearch deployment issue was found:" + "\n-------\n" + "{}".format(check_error)) + return {"failed": True, "changed": False, "msg": msg} + + # TODO(lmeyer): run it all again for the ops cluster + return {"failed": False, "changed": False, "msg": 'No problems found with Elasticsearch deployment.'} + + def _not_running_elasticsearch_pods(self, es_pods): + """Returns: list of running pods, list of errors about non-running pods""" + not_running = super(Elasticsearch, self).not_running_pods(es_pods) + if not_running: + return not_running, [( + 'The following Elasticsearch pods are not running:\n' + '{pods}' + 'These pods will not aggregate logs from their nodes.' + ).format(pods=''.join( + " {} ({})\n".format(pod['metadata']['name'], pod['spec'].get('host', 'None')) + for pod in not_running + ))] + return not_running, [] + + def check_elasticsearch(self, es_pods, task_vars): + """Various checks for elasticsearch. Returns: error string""" + not_running_pods, error_msgs = self._not_running_elasticsearch_pods(es_pods) + running_pods = [pod for pod in es_pods if pod not in not_running_pods] + pods_by_name = { + pod['metadata']['name']: pod for pod in running_pods + # Filter out pods that are not members of a DC + if pod['metadata'].get('labels', {}).get('deploymentconfig') + } + if not pods_by_name: + return 'No logging Elasticsearch pods were found. Is logging deployed?' + error_msgs += self._check_elasticsearch_masters(pods_by_name, task_vars) + error_msgs += self._check_elasticsearch_node_list(pods_by_name, task_vars) + error_msgs += self._check_es_cluster_health(pods_by_name, task_vars) + error_msgs += self._check_elasticsearch_diskspace(pods_by_name, task_vars) + return '\n'.join(error_msgs) + + @staticmethod + def _build_es_curl_cmd(pod_name, url): + base = "exec {name} -- curl -s --cert {base}cert --key {base}key --cacert {base}ca -XGET '{url}'" + return base.format(base="/etc/elasticsearch/secret/admin-", name=pod_name, url=url) + + def _check_elasticsearch_masters(self, pods_by_name, task_vars): + """Check that Elasticsearch masters are sane. Returns: list of error strings""" + es_master_names = set() + error_msgs = [] + for pod_name in pods_by_name.keys(): + # Compare what each ES node reports as master and compare for split brain + get_master_cmd = self._build_es_curl_cmd(pod_name, "https://localhost:9200/_cat/master") + master_name_str = self._exec_oc(get_master_cmd, [], task_vars) + master_names = (master_name_str or '').split(' ') + if len(master_names) > 1: + es_master_names.add(master_names[1]) + else: + error_msgs.append( + 'No master? Elasticsearch {pod} returned bad string when asked master name:\n' + ' {response}'.format(pod=pod_name, response=master_name_str) + ) + + if not es_master_names: + error_msgs.append('No logging Elasticsearch masters were found. Is logging deployed?') + return '\n'.join(error_msgs) + + if len(es_master_names) > 1: + error_msgs.append( + 'Found multiple Elasticsearch masters according to the pods:\n' + '{master_list}\n' + 'This implies that the masters have "split brain" and are not correctly\n' + 'replicating data for the logging cluster. Log loss is likely to occur.' + .format(master_list='\n'.join(' ' + master for master in es_master_names)) + ) + + return error_msgs + + def _check_elasticsearch_node_list(self, pods_by_name, task_vars): + """Check that reported ES masters are accounted for by pods. Returns: list of error strings""" + + if not pods_by_name: + return ['No logging Elasticsearch masters were found. Is logging deployed?'] + + # get ES cluster nodes + node_cmd = self._build_es_curl_cmd(list(pods_by_name.keys())[0], 'https://localhost:9200/_nodes') + cluster_node_data = self._exec_oc(node_cmd, [], task_vars) + try: + cluster_nodes = json.loads(cluster_node_data)['nodes'] + except (ValueError, KeyError): + return [ + 'Failed to query Elasticsearch for the list of ES nodes. The output was:\n' + + cluster_node_data + ] + + # Try to match all ES-reported node hosts to known pods. + error_msgs = [] + for node in cluster_nodes.values(): + # Note that with 1.4/3.4 the pod IP may be used as the master name + if not any(node['host'] in (pod_name, pod['status'].get('podIP')) + for pod_name, pod in pods_by_name.items()): + error_msgs.append( + 'The Elasticsearch cluster reports a member node "{node}"\n' + 'that does not correspond to any known ES pod.'.format(node=node['host']) + ) + + return error_msgs + + def _check_es_cluster_health(self, pods_by_name, task_vars): + """Exec into the elasticsearch pods and check the cluster health. Returns: list of errors""" + error_msgs = [] + for pod_name in pods_by_name.keys(): + cluster_health_cmd = self._build_es_curl_cmd(pod_name, 'https://localhost:9200/_cluster/health?pretty=true') + cluster_health_data = self._exec_oc(cluster_health_cmd, [], task_vars) + try: + health_res = json.loads(cluster_health_data) + if not health_res or not health_res.get('status'): + raise ValueError() + except ValueError: + error_msgs.append( + 'Could not retrieve cluster health status from logging ES pod "{pod}".\n' + 'Response was:\n{output}'.format(pod=pod_name, output=cluster_health_data) + ) + continue + + if health_res['status'] not in ['green', 'yellow']: + error_msgs.append( + 'Elasticsearch cluster health status is RED according to pod "{}"'.format(pod_name) + ) + + return error_msgs + + def _check_elasticsearch_diskspace(self, pods_by_name, task_vars): + """ + Exec into an ES pod and query the diskspace on the persistent volume. + Returns: list of errors + """ + error_msgs = [] + for pod_name in pods_by_name.keys(): + df_cmd = 'exec {} -- df --output=ipcent,pcent /elasticsearch/persistent'.format(pod_name) + disk_output = self._exec_oc(df_cmd, [], task_vars) + lines = disk_output.splitlines() + # expecting one header looking like 'IUse% Use%' and one body line + body_re = r'\s*(\d+)%?\s+(\d+)%?\s*$' + if len(lines) != 2 or len(lines[0].split()) != 2 or not re.match(body_re, lines[1]): + error_msgs.append( + 'Could not retrieve storage usage from logging ES pod "{pod}".\n' + 'Response to `df` command was:\n{output}'.format(pod=pod_name, output=disk_output) + ) + continue + inode_pct, disk_pct = re.match(body_re, lines[1]).groups() + + inode_pct_thresh = get_var(task_vars, 'openshift_check_efk_es_inode_pct', default='90') + if int(inode_pct) >= int(inode_pct_thresh): + error_msgs.append( + 'Inode percent usage on the storage volume for logging ES pod "{pod}"\n' + ' is {pct}, greater than threshold {limit}.\n' + ' Note: threshold can be specified in inventory with {param}'.format( + pod=pod_name, + pct=str(inode_pct), + limit=str(inode_pct_thresh), + param='openshift_check_efk_es_inode_pct', + )) + disk_pct_thresh = get_var(task_vars, 'openshift_check_efk_es_storage_pct', default='80') + if int(disk_pct) >= int(disk_pct_thresh): + error_msgs.append( + 'Disk percent usage on the storage volume for logging ES pod "{pod}"\n' + ' is {pct}, greater than threshold {limit}.\n' + ' Note: threshold can be specified in inventory with {param}'.format( + pod=pod_name, + pct=str(disk_pct), + limit=str(disk_pct_thresh), + param='openshift_check_efk_es_storage_pct', + )) + + return error_msgs + + def _exec_oc(self, cmd_str, extra_args, task_vars): + return super(Elasticsearch, self).exec_oc( + self.execute_module, + self.logging_namespace, + cmd_str, + extra_args, + task_vars, + ) diff --git a/roles/openshift_health_checker/openshift_checks/logging/fluentd.py b/roles/openshift_health_checker/openshift_checks/logging/fluentd.py new file mode 100644 index 000000000..627567293 --- /dev/null +++ b/roles/openshift_health_checker/openshift_checks/logging/fluentd.py @@ -0,0 +1,170 @@ +""" +Module for performing checks on an Fluentd logging deployment +""" + +import json + +from openshift_checks import get_var +from openshift_checks.logging.logging import LoggingCheck + + +class Fluentd(LoggingCheck): + """Module that checks an integrated logging Fluentd deployment""" + name = "fluentd" + tags = ["health", "logging"] + + logging_namespace = None + + def run(self, tmp, task_vars): + """Check various things and gather errors. Returns: result as hash""" + + self.logging_namespace = get_var(task_vars, "openshift_logging_namespace", default="logging") + fluentd_pods, error = super(Fluentd, self).get_pods_for_component( + self.execute_module, + self.logging_namespace, + "fluentd", + task_vars, + ) + if error: + return {"failed": True, "changed": False, "msg": error} + check_error = self.check_fluentd(fluentd_pods, task_vars) + + if check_error: + msg = ("The following Fluentd deployment issue was found:" + "\n-------\n" + "{}".format(check_error)) + return {"failed": True, "changed": False, "msg": msg} + + # TODO(lmeyer): run it all again for the ops cluster + return {"failed": False, "changed": False, "msg": 'No problems found with Fluentd deployment.'} + + @staticmethod + def _filter_fluentd_labeled_nodes(nodes_by_name, node_selector): + """Filter to all nodes with fluentd label. Returns dict(name: node), error string""" + label, value = node_selector.split('=', 1) + fluentd_nodes = { + name: node for name, node in nodes_by_name.items() + if node['metadata']['labels'].get(label) == value + } + if not fluentd_nodes: + return None, ( + 'There are no nodes with the fluentd label {label}.\n' + 'This means no logs will be aggregated from the nodes.' + ).format(label=node_selector) + return fluentd_nodes, None + + @staticmethod + def _check_node_labeling(nodes_by_name, fluentd_nodes, node_selector, task_vars): + """Note if nodes are not labeled as expected. Returns: error string""" + intended_nodes = get_var(task_vars, 'openshift_logging_fluentd_hosts', default=['--all']) + if not intended_nodes or '--all' in intended_nodes: + intended_nodes = nodes_by_name.keys() + nodes_missing_labels = set(intended_nodes) - set(fluentd_nodes.keys()) + if nodes_missing_labels: + return ( + 'The following nodes are supposed to be labeled with {label} but are not:\n' + ' {nodes}\n' + 'Fluentd will not aggregate logs from these nodes.' + ).format(label=node_selector, nodes=', '.join(nodes_missing_labels)) + return None + + @staticmethod + def _check_nodes_have_fluentd(pods, fluentd_nodes): + """Make sure fluentd is on all the labeled nodes. Returns: error string""" + unmatched_nodes = fluentd_nodes.copy() + node_names_by_label = { + node['metadata']['labels']['kubernetes.io/hostname']: name + for name, node in fluentd_nodes.items() + } + node_names_by_internal_ip = { + address['address']: name + for name, node in fluentd_nodes.items() + for address in node['status']['addresses'] + if address['type'] == "InternalIP" + } + for pod in pods: + for name in [ + pod['spec']['nodeName'], + node_names_by_internal_ip.get(pod['spec']['nodeName']), + node_names_by_label.get(pod.get('spec', {}).get('host')), + ]: + unmatched_nodes.pop(name, None) + if unmatched_nodes: + return ( + 'The following nodes are supposed to have a Fluentd pod but do not:\n' + '{nodes}' + 'These nodes will not have their logs aggregated.' + ).format(nodes=''.join( + " {}\n".format(name) + for name in unmatched_nodes.keys() + )) + return None + + def _check_fluentd_pods_running(self, pods): + """Make sure all fluentd pods are running. Returns: error string""" + not_running = super(Fluentd, self).not_running_pods(pods) + if not_running: + return ( + 'The following Fluentd pods are supposed to be running but are not:\n' + '{pods}' + 'These pods will not aggregate logs from their nodes.' + ).format(pods=''.join( + " {} ({})\n".format(pod['metadata']['name'], pod['spec'].get('host', 'None')) + for pod in not_running + )) + return None + + def check_fluentd(self, pods, task_vars): + """Verify fluentd is running everywhere. Returns: error string""" + + node_selector = get_var(task_vars, 'openshift_logging_fluentd_nodeselector', + default='logging-infra-fluentd=true') + + nodes_by_name, error = self.get_nodes_by_name(task_vars) + + if error: + return error + fluentd_nodes, error = self._filter_fluentd_labeled_nodes(nodes_by_name, node_selector) + if error: + return error + + error_msgs = [] + error = self._check_node_labeling(nodes_by_name, fluentd_nodes, node_selector, task_vars) + if error: + error_msgs.append(error) + error = self._check_nodes_have_fluentd(pods, fluentd_nodes) + if error: + error_msgs.append(error) + error = self._check_fluentd_pods_running(pods) + if error: + error_msgs.append(error) + + # Make sure there are no extra fluentd pods + if len(pods) > len(fluentd_nodes): + error_msgs.append( + 'There are more Fluentd pods running than nodes labeled.\n' + 'This may not cause problems with logging but it likely indicates something wrong.' + ) + + return '\n'.join(error_msgs) + + def get_nodes_by_name(self, task_vars): + """Retrieve all the node definitions. Returns: dict(name: node), error string""" + nodes_json = self._exec_oc("get nodes -o json", [], task_vars) + try: + nodes = json.loads(nodes_json) + except ValueError: # no valid json - should not happen + return None, "Could not obtain a list of nodes to validate fluentd. Output from oc get:\n" + nodes_json + if not nodes or not nodes.get('items'): # also should not happen + return None, "No nodes appear to be defined according to the API." + return { + node['metadata']['name']: node + for node in nodes['items'] + }, None + + def _exec_oc(self, cmd_str, extra_args, task_vars): + return super(Fluentd, self).exec_oc(self.execute_module, + self.logging_namespace, + cmd_str, + extra_args, + task_vars) diff --git a/roles/openshift_health_checker/openshift_checks/logging/kibana.py b/roles/openshift_health_checker/openshift_checks/logging/kibana.py new file mode 100644 index 000000000..442f407b1 --- /dev/null +++ b/roles/openshift_health_checker/openshift_checks/logging/kibana.py @@ -0,0 +1,229 @@ +""" +Module for performing checks on a Kibana logging deployment +""" + +import json +import ssl + +try: + from urllib2 import HTTPError, URLError + import urllib2 +except ImportError: + from urllib.error import HTTPError, URLError + import urllib.request as urllib2 + +from openshift_checks import get_var +from openshift_checks.logging.logging import LoggingCheck + + +class Kibana(LoggingCheck): + """Module that checks an integrated logging Kibana deployment""" + + name = "kibana" + tags = ["health", "logging"] + + logging_namespace = None + + def run(self, tmp, task_vars): + """Check various things and gather errors. Returns: result as hash""" + + self.logging_namespace = get_var(task_vars, "openshift_logging_namespace", default="logging") + kibana_pods, error = super(Kibana, self).get_pods_for_component( + self.execute_module, + self.logging_namespace, + "kibana", + task_vars, + ) + if error: + return {"failed": True, "changed": False, "msg": error} + check_error = self.check_kibana(kibana_pods) + + if not check_error: + check_error = self._check_kibana_route(task_vars) + + if check_error: + msg = ("The following Kibana deployment issue was found:" + "\n-------\n" + "{}".format(check_error)) + return {"failed": True, "changed": False, "msg": msg} + + # TODO(lmeyer): run it all again for the ops cluster + return {"failed": False, "changed": False, "msg": 'No problems found with Kibana deployment.'} + + def _verify_url_internal(self, url, task_vars): + """ + Try to reach a URL from the host. + Returns: success (bool), reason (for failure) + """ + args = dict( + url=url, + follow_redirects='none', + validate_certs='no', # likely to be signed with internal CA + # TODO(lmeyer): give users option to validate certs + status_code=302, + ) + result = self.execute_module('uri', args, task_vars) + if result.get('failed'): + return result['msg'] + return None + + @staticmethod + def _verify_url_external(url): + """ + Try to reach a URL from ansible control host. + Returns: success (bool), reason (for failure) + """ + # This actually checks from the ansible control host, which may or may not + # really be "external" to the cluster. + + # Disable SSL cert validation to work around internally signed certs + ctx = ssl.create_default_context() + ctx.check_hostname = False # or setting CERT_NONE is refused + ctx.verify_mode = ssl.CERT_NONE + + # Verify that the url is returning a valid response + try: + # We only care if the url connects and responds + return_code = urllib2.urlopen(url, context=ctx).getcode() + except HTTPError as httperr: + return httperr.reason + except URLError as urlerr: + return str(urlerr) + + # there appears to be no way to prevent urlopen from following redirects + if return_code != 200: + return 'Expected success (200) but got return code {}'.format(int(return_code)) + + return None + + def check_kibana(self, pods): + """Check to see if Kibana is up and working. Returns: error string.""" + + if not pods: + return "There are no Kibana pods deployed, so no access to the logging UI." + + not_running = self.not_running_pods(pods) + if len(not_running) == len(pods): + return "No Kibana pod is in a running state, so there is no access to the logging UI." + elif not_running: + return ( + "The following Kibana pods are not currently in a running state:\n" + "{pods}" + "However at least one is, so service may not be impacted." + ).format(pods="".join(" " + pod['metadata']['name'] + "\n" for pod in not_running)) + + return None + + def _get_kibana_url(self, task_vars): + """ + Get kibana route or report error. + Returns: url (or empty), reason for failure + """ + + # Get logging url + get_route = self._exec_oc("get route logging-kibana -o json", [], task_vars) + if not get_route: + return None, 'no_route_exists' + + route = json.loads(get_route) + + # check that the route has been accepted by a router + ingress = route["status"]["ingress"] + # ingress can be null if there is no router, or empty if not routed + if not ingress or not ingress[0]: + return None, 'route_not_accepted' + + host = route.get("spec", {}).get("host") + if not host: + return None, 'route_missing_host' + + return 'https://{}/'.format(host), None + + def _check_kibana_route(self, task_vars): + """ + Check to see if kibana route is up and working. + Returns: error string + """ + known_errors = dict( + no_route_exists=( + 'No route is defined for Kibana in the logging namespace,\n' + 'so the logging stack is not accessible. Is logging deployed?\n' + 'Did something remove the logging-kibana route?' + ), + route_not_accepted=( + 'The logging-kibana route is not being routed by any router.\n' + 'Is the router deployed and working?' + ), + route_missing_host=( + 'The logging-kibana route has no hostname defined,\n' + 'which should never happen. Did something alter its definition?' + ), + ) + + kibana_url, error = self._get_kibana_url(task_vars) + if not kibana_url: + return known_errors.get(error, error) + + # first, check that kibana is reachable from the master. + error = self._verify_url_internal(kibana_url, task_vars) + if error: + if 'urlopen error [Errno 111] Connection refused' in error: + error = ( + 'Failed to connect from this master to Kibana URL {url}\n' + 'Is kibana running, and is at least one router routing to it?' + ).format(url=kibana_url) + elif 'urlopen error [Errno -2] Name or service not known' in error: + error = ( + 'Failed to connect from this master to Kibana URL {url}\n' + 'because the hostname does not resolve.\n' + 'Is DNS configured for the Kibana hostname?' + ).format(url=kibana_url) + elif 'Status code was not' in error: + error = ( + 'A request from this master to the Kibana URL {url}\n' + 'did not return the correct status code (302).\n' + 'This could mean that Kibana is malfunctioning, the hostname is\n' + 'resolving incorrectly, or other network issues. The output was:\n' + ' {error}' + ).format(url=kibana_url, error=error) + return 'Error validating the logging Kibana route:\n' + error + + # in production we would like the kibana route to work from outside the + # cluster too; but that may not be the case, so allow disabling just this part. + if not get_var(task_vars, "openshift_check_efk_kibana_external", default=True): + return None + error = self._verify_url_external(kibana_url) + if error: + if 'urlopen error [Errno 111] Connection refused' in error: + error = ( + 'Failed to connect from the Ansible control host to Kibana URL {url}\n' + 'Is the router for the Kibana hostname exposed externally?' + ).format(url=kibana_url) + elif 'urlopen error [Errno -2] Name or service not known' in error: + error = ( + 'Failed to resolve the Kibana hostname in {url}\n' + 'from the Ansible control host.\n' + 'Is DNS configured to resolve this Kibana hostname externally?' + ).format(url=kibana_url) + elif 'Expected success (200)' in error: + error = ( + 'A request to Kibana at {url}\n' + 'returned the wrong error code:\n' + ' {error}\n' + 'This could mean that Kibana is malfunctioning, the hostname is\n' + 'resolving incorrectly, or other network issues.' + ).format(url=kibana_url, error=error) + error = ( + 'Error validating the logging Kibana route:\n{error}\n' + 'To disable external Kibana route validation, set in your inventory:\n' + ' openshift_check_efk_kibana_external=False' + ).format(error=error) + return error + return None + + def _exec_oc(self, cmd_str, extra_args, task_vars): + return super(Kibana, self).exec_oc(self.execute_module, + self.logging_namespace, + cmd_str, + extra_args, + task_vars) diff --git a/roles/openshift_health_checker/openshift_checks/logging/logging.py b/roles/openshift_health_checker/openshift_checks/logging/logging.py new file mode 100644 index 000000000..05b4d300c --- /dev/null +++ b/roles/openshift_health_checker/openshift_checks/logging/logging.py @@ -0,0 +1,96 @@ +""" +Util functions for performing checks on an Elasticsearch, Fluentd, and Kibana stack +""" + +import json +import os + +from openshift_checks import OpenShiftCheck, OpenShiftCheckException, get_var + + +class LoggingCheck(OpenShiftCheck): + """Base class for logging component checks""" + + name = "logging" + + @classmethod + def is_active(cls, task_vars): + return super(LoggingCheck, cls).is_active(task_vars) and cls.is_first_master(task_vars) + + @staticmethod + def is_first_master(task_vars): + """Run only on first master and only when logging is configured. Returns: bool""" + logging_deployed = get_var(task_vars, "openshift_hosted_logging_deploy", default=True) + # Note: It would be nice to use membership in oo_first_master group, however for now it + # seems best to avoid requiring that setup and just check this is the first master. + hostname = get_var(task_vars, "ansible_ssh_host") or [None] + masters = get_var(task_vars, "groups", "masters", default=None) or [None] + return logging_deployed and masters[0] == hostname + + def run(self, tmp, task_vars): + pass + + def get_pods_for_component(self, execute_module, namespace, logging_component, task_vars): + """Get all pods for a given component. Returns: list of pods for component, error string""" + pod_output = self.exec_oc( + execute_module, + namespace, + "get pods -l component={} -o json".format(logging_component), + [], + task_vars + ) + try: + pods = json.loads(pod_output) + if not pods or not pods.get('items'): + raise ValueError() + except ValueError: + # successful run but non-parsing data generally means there were no pods in the namespace + return None, 'There are no pods in the {} namespace. Is logging deployed?'.format(namespace) + + return pods['items'], None + + @staticmethod + def not_running_pods(pods): + """Returns: list of pods not in a ready and running state""" + return [ + pod for pod in pods + if any( + container['ready'] is False + for container in pod['status']['containerStatuses'] + ) or not any( + condition['type'] == 'Ready' and condition['status'] == 'True' + for condition in pod['status']['conditions'] + ) + ] + + @staticmethod + def exec_oc(execute_module=None, namespace="logging", cmd_str="", extra_args=None, task_vars=None): + """ + Execute an 'oc' command in the remote host. + Returns: output of command and namespace, + or raises OpenShiftCheckException on error + """ + config_base = get_var(task_vars, "openshift", "common", "config_base") + args = { + "namespace": namespace, + "config_file": os.path.join(config_base, "master", "admin.kubeconfig"), + "cmd": cmd_str, + "extra_args": list(extra_args) if extra_args else [], + } + + result = execute_module("ocutil", args, task_vars) + if result.get("failed"): + msg = ( + 'Unexpected error using `oc` to validate the logging stack components.\n' + 'Error executing `oc {cmd}`:\n' + '{error}' + ).format(cmd=args['cmd'], error=result['result']) + + if result['result'] == '[Errno 2] No such file or directory': + msg = ( + "This host is supposed to be a master but does not have the `oc` command where expected.\n" + "Has an installation been run on this host yet?" + ) + raise OpenShiftCheckException(msg) + + return result.get("result", "") diff --git a/roles/openshift_health_checker/openshift_checks/memory_availability.py b/roles/openshift_health_checker/openshift_checks/memory_availability.py new file mode 100644 index 000000000..f4e31065f --- /dev/null +++ b/roles/openshift_health_checker/openshift_checks/memory_availability.py @@ -0,0 +1,51 @@ +# pylint: disable=missing-docstring +from openshift_checks import OpenShiftCheck, get_var + +MIB = 2**20 +GIB = 2**30 + + +class MemoryAvailability(OpenShiftCheck): + """Check that recommended memory is available.""" + + name = "memory_availability" + tags = ["preflight"] + + # Values taken from the official installation documentation: + # https://docs.openshift.org/latest/install_config/install/prerequisites.html#system-requirements + recommended_memory_bytes = { + "masters": 16 * GIB, + "nodes": 8 * GIB, + "etcd": 8 * GIB, + } + # https://access.redhat.com/solutions/3006511 physical RAM is partly reserved from memtotal + memtotal_adjustment = 1 * GIB + + @classmethod + def is_active(cls, task_vars): + """Skip hosts that do not have recommended memory requirements.""" + group_names = get_var(task_vars, "group_names", default=[]) + has_memory_recommendation = bool(set(group_names).intersection(cls.recommended_memory_bytes)) + return super(MemoryAvailability, cls).is_active(task_vars) and has_memory_recommendation + + def run(self, tmp, task_vars): + group_names = get_var(task_vars, "group_names") + total_memory_bytes = get_var(task_vars, "ansible_memtotal_mb") * MIB + + recommended_min = max(self.recommended_memory_bytes.get(name, 0) for name in group_names) + configured_min = float(get_var(task_vars, "openshift_check_min_host_memory_gb", default=0)) * GIB + min_memory_bytes = configured_min or recommended_min + + if total_memory_bytes + self.memtotal_adjustment < min_memory_bytes: + return { + 'failed': True, + 'msg': ( + 'Available memory ({available:.1f} GiB) is too far ' + 'below recommended value ({recommended:.1f} GiB)' + ).format( + available=float(total_memory_bytes) / GIB, + recommended=float(min_memory_bytes) / GIB, + ), + } + + return {} diff --git a/roles/openshift_health_checker/openshift_checks/mixins.py b/roles/openshift_health_checker/openshift_checks/mixins.py index 657e15160..2cb2e21aa 100644 --- a/roles/openshift_health_checker/openshift_checks/mixins.py +++ b/roles/openshift_health_checker/openshift_checks/mixins.py @@ -1,18 +1,58 @@ -# pylint: disable=missing-docstring +""" +Mixin classes meant to be used with subclasses of OpenShiftCheck. +""" + from openshift_checks import get_var class NotContainerizedMixin(object): """Mixin for checks that are only active when not in containerized mode.""" + # permanent # pylint: disable=too-few-public-methods + # Reason: The mixin is not intended to stand on its own as a class. @classmethod def is_active(cls, task_vars): - return ( - # This mixin is meant to be used with subclasses of OpenShiftCheck. - super(NotContainerizedMixin, cls).is_active(task_vars) and - not cls.is_containerized(task_vars) - ) + """Only run on non-containerized hosts.""" + is_containerized = get_var(task_vars, "openshift", "common", "is_containerized") + return super(NotContainerizedMixin, cls).is_active(task_vars) and not is_containerized + + +class DockerHostMixin(object): + """Mixin for checks that are only active on hosts that require Docker.""" + + dependencies = [] - @staticmethod - def is_containerized(task_vars): - return get_var(task_vars, "openshift", "common", "is_containerized") + @classmethod + def is_active(cls, task_vars): + """Only run on hosts that depend on Docker.""" + is_containerized = get_var(task_vars, "openshift", "common", "is_containerized") + is_node = "nodes" in get_var(task_vars, "group_names", default=[]) + return super(DockerHostMixin, cls).is_active(task_vars) and (is_containerized or is_node) + + def ensure_dependencies(self, task_vars): + """ + Ensure that docker-related packages exist, but not on atomic hosts + (which would not be able to install but should already have them). + Returns: msg, failed, changed + """ + if get_var(task_vars, "openshift", "common", "is_atomic"): + return "", False, False + + # NOTE: we would use the "package" module but it's actually an action plugin + # and it's not clear how to invoke one of those. This is about the same anyway: + result = self.execute_module( + get_var(task_vars, "ansible_pkg_mgr", default="yum"), + {"name": self.dependencies, "state": "present"}, + task_vars=task_vars, + ) + msg = result.get("msg", "") + if result.get("failed"): + if "No package matching" in msg: + msg = "Ensure that all required dependencies can be installed via `yum`.\n" + msg = ( + "Unable to install required packages on this host:\n" + " {deps}\n{msg}" + ).format(deps=',\n '.join(self.dependencies), msg=msg) + failed = result.get("failed", False) or result.get("rc", 0) != 0 + changed = result.get("changed", False) + return msg, failed, changed diff --git a/roles/openshift_health_checker/openshift_checks/ovs_version.py b/roles/openshift_health_checker/openshift_checks/ovs_version.py new file mode 100644 index 000000000..2dd045f1f --- /dev/null +++ b/roles/openshift_health_checker/openshift_checks/ovs_version.py @@ -0,0 +1,78 @@ +""" +Ansible module for determining if an installed version of Open vSwitch is incompatible with the +currently installed version of OpenShift. +""" + +from openshift_checks import OpenShiftCheck, OpenShiftCheckException, get_var +from openshift_checks.mixins import NotContainerizedMixin + + +class OvsVersion(NotContainerizedMixin, OpenShiftCheck): + """Check that packages in a package_list are installed on the host + and are the correct version as determined by an OpenShift installation. + """ + + name = "ovs_version" + tags = ["health"] + + openshift_to_ovs_version = { + "3.6": "2.6", + "3.5": "2.6", + "3.4": "2.4", + } + + # map major release versions across releases + # to a common major version + openshift_major_release_version = { + "1": "3", + } + + @classmethod + def is_active(cls, task_vars): + """Skip hosts that do not have package requirements.""" + group_names = get_var(task_vars, "group_names", default=[]) + master_or_node = 'masters' in group_names or 'nodes' in group_names + return super(OvsVersion, cls).is_active(task_vars) and master_or_node + + def run(self, tmp, task_vars): + args = { + "package_list": [ + { + "name": "openvswitch", + "version": self.get_required_ovs_version(task_vars), + }, + ], + } + return self.execute_module("rpm_version", args, task_vars=task_vars) + + def get_required_ovs_version(self, task_vars): + """Return the correct Open vSwitch version for the current OpenShift version""" + openshift_version = self._get_openshift_version(task_vars) + + if float(openshift_version) < 3.5: + return self.openshift_to_ovs_version["3.4"] + + ovs_version = self.openshift_to_ovs_version.get(str(openshift_version)) + if ovs_version: + return self.openshift_to_ovs_version[str(openshift_version)] + + msg = "There is no recommended version of Open vSwitch for the current version of OpenShift: {}" + raise OpenShiftCheckException(msg.format(openshift_version)) + + def _get_openshift_version(self, task_vars): + openshift_version = get_var(task_vars, "openshift_image_tag") + if openshift_version and openshift_version[0] == 'v': + openshift_version = openshift_version[1:] + + return self._parse_version(openshift_version) + + def _parse_version(self, version): + components = version.split(".") + if not components or len(components) < 2: + msg = "An invalid version of OpenShift was found for this host: {}" + raise OpenShiftCheckException(msg.format(version)) + + if components[0] in self.openshift_major_release_version: + components[0] = self.openshift_major_release_version[components[0]] + + return '.'.join(components[:2]) diff --git a/roles/openshift_health_checker/openshift_checks/package_availability.py b/roles/openshift_health_checker/openshift_checks/package_availability.py index 9891972a6..0dd2b1286 100644 --- a/roles/openshift_health_checker/openshift_checks/package_availability.py +++ b/roles/openshift_health_checker/openshift_checks/package_availability.py @@ -9,6 +9,10 @@ class PackageAvailability(NotContainerizedMixin, OpenShiftCheck): name = "package_availability" tags = ["preflight"] + @classmethod + def is_active(cls, task_vars): + return super(PackageAvailability, cls).is_active(task_vars) and task_vars["ansible_pkg_mgr"] == "yum" + def run(self, tmp, task_vars): rpm_prefix = get_var(task_vars, "openshift", "common", "service_type") group_names = get_var(task_vars, "group_names", default=[]) @@ -21,7 +25,7 @@ class PackageAvailability(NotContainerizedMixin, OpenShiftCheck): packages.update(self.node_packages(rpm_prefix)) args = {"packages": sorted(set(packages))} - return self.execute_module("check_yum_update", args, tmp, task_vars) + return self.execute_module("check_yum_update", args, tmp=tmp, task_vars=task_vars) @staticmethod def master_packages(rpm_prefix): @@ -32,8 +36,7 @@ class PackageAvailability(NotContainerizedMixin, OpenShiftCheck): "bash-completion", "cockpit-bridge", "cockpit-docker", - "cockpit-kubernetes", - "cockpit-shell", + "cockpit-system", "cockpit-ws", "etcd", "httpd-tools", diff --git a/roles/openshift_health_checker/openshift_checks/package_update.py b/roles/openshift_health_checker/openshift_checks/package_update.py index fd0c0a755..f432380c6 100644 --- a/roles/openshift_health_checker/openshift_checks/package_update.py +++ b/roles/openshift_health_checker/openshift_checks/package_update.py @@ -11,4 +11,4 @@ class PackageUpdate(NotContainerizedMixin, OpenShiftCheck): def run(self, tmp, task_vars): args = {"packages": []} - return self.execute_module("check_yum_update", args, tmp, task_vars) + return self.execute_module("check_yum_update", args, tmp=tmp, task_vars=task_vars) diff --git a/roles/openshift_health_checker/openshift_checks/package_version.py b/roles/openshift_health_checker/openshift_checks/package_version.py index 42193a1c6..6a76bb93d 100644 --- a/roles/openshift_health_checker/openshift_checks/package_version.py +++ b/roles/openshift_health_checker/openshift_checks/package_version.py @@ -1,5 +1,5 @@ # pylint: disable=missing-docstring -from openshift_checks import OpenShiftCheck, get_var +from openshift_checks import OpenShiftCheck, OpenShiftCheckException, get_var from openshift_checks.mixins import NotContainerizedMixin @@ -9,12 +9,117 @@ class PackageVersion(NotContainerizedMixin, OpenShiftCheck): name = "package_version" tags = ["preflight"] + openshift_to_ovs_version = { + "3.6": "2.6", + "3.5": "2.6", + "3.4": "2.4", + } + + openshift_to_docker_version = { + "3.1": "1.8", + "3.2": "1.10", + "3.3": "1.10", + "3.4": "1.12", + } + + # map major release versions across releases + # to a common major version + openshift_major_release_version = { + "1": "3", + } + + @classmethod + def is_active(cls, task_vars): + """Skip hosts that do not have package requirements.""" + group_names = get_var(task_vars, "group_names", default=[]) + master_or_node = 'masters' in group_names or 'nodes' in group_names + return super(PackageVersion, cls).is_active(task_vars) and master_or_node + def run(self, tmp, task_vars): rpm_prefix = get_var(task_vars, "openshift", "common", "service_type") - openshift_release = get_var(task_vars, "openshift_release") + openshift_release = get_var(task_vars, "openshift_release", default='') + deployment_type = get_var(task_vars, "openshift_deployment_type") + check_multi_minor_release = deployment_type in ['openshift-enterprise'] args = { - "prefix": rpm_prefix, - "version": openshift_release, + "package_list": [ + { + "name": "openvswitch", + "version": self.get_required_ovs_version(task_vars), + "check_multi": False, + }, + { + "name": "docker", + "version": self.get_required_docker_version(task_vars), + "check_multi": False, + }, + { + "name": "{}".format(rpm_prefix), + "version": openshift_release, + "check_multi": check_multi_minor_release, + }, + { + "name": "{}-master".format(rpm_prefix), + "version": openshift_release, + "check_multi": check_multi_minor_release, + }, + { + "name": "{}-node".format(rpm_prefix), + "version": openshift_release, + "check_multi": check_multi_minor_release, + }, + ], } - return self.execute_module("aos_version", args, tmp, task_vars) + + return self.execute_module("aos_version", args, tmp=tmp, task_vars=task_vars) + + def get_required_ovs_version(self, task_vars): + """Return the correct Open vSwitch version for the current OpenShift version. + If the current OpenShift version is >= 3.5, ensure Open vSwitch version 2.6, + Else ensure Open vSwitch version 2.4""" + openshift_version = self.get_openshift_version(task_vars) + + if float(openshift_version) < 3.5: + return self.openshift_to_ovs_version["3.4"] + + ovs_version = self.openshift_to_ovs_version.get(str(openshift_version)) + if ovs_version: + return ovs_version + + msg = "There is no recommended version of Open vSwitch for the current version of OpenShift: {}" + raise OpenShiftCheckException(msg.format(openshift_version)) + + def get_required_docker_version(self, task_vars): + """Return the correct Docker version for the current OpenShift version. + If the OpenShift version is 3.1, ensure Docker version 1.8. + If the OpenShift version is 3.2 or 3.3, ensure Docker version 1.10. + If the current OpenShift version is >= 3.4, ensure Docker version 1.12.""" + openshift_version = self.get_openshift_version(task_vars) + + if float(openshift_version) >= 3.4: + return self.openshift_to_docker_version["3.4"] + + docker_version = self.openshift_to_docker_version.get(str(openshift_version)) + if docker_version: + return docker_version + + msg = "There is no recommended version of Docker for the current version of OpenShift: {}" + raise OpenShiftCheckException(msg.format(openshift_version)) + + def get_openshift_version(self, task_vars): + openshift_version = get_var(task_vars, "openshift_image_tag") + if openshift_version and openshift_version[0] == 'v': + openshift_version = openshift_version[1:] + + return self.parse_version(openshift_version) + + def parse_version(self, version): + components = version.split(".") + if not components or len(components) < 2: + msg = "An invalid version of OpenShift was found for this host: {}" + raise OpenShiftCheckException(msg.format(version)) + + if components[0] in self.openshift_major_release_version: + components[0] = self.openshift_major_release_version[components[0]] + + return '.'.join(components[:2]) diff --git a/roles/openshift_health_checker/test/action_plugin_test.py b/roles/openshift_health_checker/test/action_plugin_test.py new file mode 100644 index 000000000..9383b233c --- /dev/null +++ b/roles/openshift_health_checker/test/action_plugin_test.py @@ -0,0 +1,253 @@ +import pytest + +from ansible.playbook.play_context import PlayContext + +from openshift_health_check import ActionModule, resolve_checks +from openshift_checks import OpenShiftCheckException + + +def fake_check(name='fake_check', tags=None, is_active=True, run_return=None, run_exception=None): + """Returns a new class that is compatible with OpenShiftCheck for testing.""" + + _name, _tags = name, tags + + class FakeCheck(object): + name = _name + tags = _tags or [] + + def __init__(self, execute_module=None): + pass + + @classmethod + def is_active(cls, task_vars): + return is_active + + def run(self, tmp, task_vars): + if run_exception is not None: + raise run_exception + return run_return + + return FakeCheck + + +# Fixtures + + +@pytest.fixture +def plugin(): + task = FakeTask('openshift_health_check', {'checks': ['fake_check']}) + plugin = ActionModule(task, None, PlayContext(), None, None, None) + return plugin + + +class FakeTask(object): + def __init__(self, action, args): + self.action = action + self.args = args + self.async = 0 + + +@pytest.fixture +def task_vars(): + return dict(openshift=dict(), ansible_host='unit-test-host') + + +# Assertion helpers + + +def failed(result, msg_has=None): + if msg_has is not None: + assert 'msg' in result + for term in msg_has: + assert term.lower() in result['msg'].lower() + return result.get('failed', False) + + +def changed(result): + return result.get('changed', False) + + +# tests whether task is skipped, not individual checks +def skipped(result): + return result.get('skipped', False) + + +# Tests + + +@pytest.mark.parametrize('task_vars', [ + None, + {}, +]) +def test_action_plugin_missing_openshift_facts(plugin, task_vars): + result = plugin.run(tmp=None, task_vars=task_vars) + + assert failed(result, msg_has=['openshift_facts']) + + +def test_action_plugin_cannot_load_checks_with_the_same_name(plugin, task_vars, monkeypatch): + FakeCheck1 = fake_check('duplicate_name') + FakeCheck2 = fake_check('duplicate_name') + checks = [FakeCheck1, FakeCheck2] + monkeypatch.setattr('openshift_checks.OpenShiftCheck.subclasses', classmethod(lambda cls: checks)) + + result = plugin.run(tmp=None, task_vars=task_vars) + + assert failed(result, msg_has=['unique', 'duplicate_name', 'FakeCheck']) + + +def test_action_plugin_skip_non_active_checks(plugin, task_vars, monkeypatch): + checks = [fake_check(is_active=False)] + monkeypatch.setattr('openshift_checks.OpenShiftCheck.subclasses', classmethod(lambda cls: checks)) + + result = plugin.run(tmp=None, task_vars=task_vars) + + assert result['checks']['fake_check'] == dict(skipped=True, skipped_reason="Not active for this host") + assert not failed(result) + assert not changed(result) + assert not skipped(result) + + +def test_action_plugin_skip_disabled_checks(plugin, task_vars, monkeypatch): + checks = [fake_check('fake_check', is_active=True)] + monkeypatch.setattr('openshift_checks.OpenShiftCheck.subclasses', classmethod(lambda cls: checks)) + + task_vars['openshift_disable_check'] = 'fake_check' + result = plugin.run(tmp=None, task_vars=task_vars) + + assert result['checks']['fake_check'] == dict(skipped=True, skipped_reason="Disabled by user request") + assert not failed(result) + assert not changed(result) + assert not skipped(result) + + +def test_action_plugin_run_check_ok(plugin, task_vars, monkeypatch): + check_return_value = {'ok': 'test'} + check_class = fake_check(run_return=check_return_value) + monkeypatch.setattr(plugin, 'load_known_checks', lambda: {'fake_check': check_class()}) + monkeypatch.setattr('openshift_health_check.resolve_checks', lambda *args: ['fake_check']) + + result = plugin.run(tmp=None, task_vars=task_vars) + + assert result['checks']['fake_check'] == check_return_value + assert not failed(result) + assert not changed(result) + assert not skipped(result) + + +def test_action_plugin_run_check_changed(plugin, task_vars, monkeypatch): + check_return_value = {'ok': 'test', 'changed': True} + check_class = fake_check(run_return=check_return_value) + monkeypatch.setattr(plugin, 'load_known_checks', lambda: {'fake_check': check_class()}) + monkeypatch.setattr('openshift_health_check.resolve_checks', lambda *args: ['fake_check']) + + result = plugin.run(tmp=None, task_vars=task_vars) + + assert result['checks']['fake_check'] == check_return_value + assert not failed(result) + assert changed(result) + assert not skipped(result) + + +def test_action_plugin_run_check_fail(plugin, task_vars, monkeypatch): + check_return_value = {'failed': True} + check_class = fake_check(run_return=check_return_value) + monkeypatch.setattr(plugin, 'load_known_checks', lambda: {'fake_check': check_class()}) + monkeypatch.setattr('openshift_health_check.resolve_checks', lambda *args: ['fake_check']) + + result = plugin.run(tmp=None, task_vars=task_vars) + + assert result['checks']['fake_check'] == check_return_value + assert failed(result, msg_has=['failed']) + assert not changed(result) + assert not skipped(result) + + +def test_action_plugin_run_check_exception(plugin, task_vars, monkeypatch): + exception_msg = 'fake check has an exception' + run_exception = OpenShiftCheckException(exception_msg) + check_class = fake_check(run_exception=run_exception) + monkeypatch.setattr(plugin, 'load_known_checks', lambda: {'fake_check': check_class()}) + monkeypatch.setattr('openshift_health_check.resolve_checks', lambda *args: ['fake_check']) + + result = plugin.run(tmp=None, task_vars=task_vars) + + assert failed(result['checks']['fake_check'], msg_has=exception_msg) + assert failed(result, msg_has=['failed']) + assert not changed(result) + assert not skipped(result) + + +def test_action_plugin_resolve_checks_exception(plugin, task_vars, monkeypatch): + monkeypatch.setattr(plugin, 'load_known_checks', lambda: {}) + + result = plugin.run(tmp=None, task_vars=task_vars) + + assert failed(result, msg_has=['unknown', 'name']) + assert not changed(result) + assert not skipped(result) + + +@pytest.mark.parametrize('names,all_checks,expected', [ + ([], [], set()), + ( + ['a', 'b'], + [ + fake_check('a'), + fake_check('b'), + ], + set(['a', 'b']), + ), + ( + ['a', 'b', '@group'], + [ + fake_check('from_group_1', ['group', 'another_group']), + fake_check('not_in_group', ['another_group']), + fake_check('from_group_2', ['preflight', 'group']), + fake_check('a'), + fake_check('b'), + ], + set(['a', 'b', 'from_group_1', 'from_group_2']), + ), +]) +def test_resolve_checks_ok(names, all_checks, expected): + assert resolve_checks(names, all_checks) == expected + + +@pytest.mark.parametrize('names,all_checks,words_in_exception,words_not_in_exception', [ + ( + ['testA', 'testB'], + [], + ['check', 'name', 'testA', 'testB'], + ['tag', 'group', '@'], + ), + ( + ['@group'], + [], + ['tag', 'name', 'group'], + ['check', '@'], + ), + ( + ['testA', 'testB', '@group'], + [], + ['check', 'name', 'testA', 'testB', 'tag', 'group'], + ['@'], + ), + ( + ['testA', 'testB', '@group'], + [ + fake_check('from_group_1', ['group', 'another_group']), + fake_check('not_in_group', ['another_group']), + fake_check('from_group_2', ['preflight', 'group']), + ], + ['check', 'name', 'testA', 'testB'], + ['tag', 'group', '@'], + ), +]) +def test_resolve_checks_failure(names, all_checks, words_in_exception, words_not_in_exception): + with pytest.raises(Exception) as excinfo: + resolve_checks(names, all_checks) + for word in words_in_exception: + assert word in str(excinfo.value) + for word in words_not_in_exception: + assert word not in str(excinfo.value) diff --git a/roles/openshift_health_checker/test/aos_version_test.py b/roles/openshift_health_checker/test/aos_version_test.py new file mode 100644 index 000000000..697805dd2 --- /dev/null +++ b/roles/openshift_health_checker/test/aos_version_test.py @@ -0,0 +1,149 @@ +import pytest +import aos_version + +from collections import namedtuple +Package = namedtuple('Package', ['name', 'version']) + +expected_pkgs = { + "spam": { + "name": "spam", + "version": "3.2.1", + "check_multi": False, + }, + "eggs": { + "name": "eggs", + "version": "3.2.1", + "check_multi": False, + }, +} + + +@pytest.mark.parametrize('pkgs, expect_not_found', [ + ( + [], + { + "spam": { + "name": "spam", + "version": "3.2.1", + "check_multi": False, + }, + "eggs": { + "name": "eggs", + "version": "3.2.1", + "check_multi": False, + } + }, # none found + ), + ( + [Package('spam', '3.2.1')], + { + "eggs": { + "name": "eggs", + "version": "3.2.1", + "check_multi": False, + } + }, # completely missing + ), + ( + [Package('spam', '3.2.1'), Package('eggs', '3.3.2')], + { + "eggs": { + "name": "eggs", + "version": "3.2.1", + "check_multi": False, + } + }, # not the right version + ), + ( + [Package('spam', '3.2.1'), Package('eggs', '3.2.1')], + {}, # all found + ), + ( + [Package('spam', '3.2.1'), Package('eggs', '3.2.1.5')], + {}, # found with more specific version + ), + ( + [Package('eggs', '1.2.3'), Package('eggs', '3.2.1.5')], + { + "spam": { + "name": "spam", + "version": "3.2.1", + "check_multi": False, + } + }, # eggs found with multiple versions + ), +]) +def test_check_pkgs_for_precise_version(pkgs, expect_not_found): + if expect_not_found: + with pytest.raises(aos_version.PreciseVersionNotFound) as e: + aos_version._check_precise_version_found(pkgs, expected_pkgs) + + assert list(expect_not_found.values()) == e.value.problem_pkgs + else: + aos_version._check_precise_version_found(pkgs, expected_pkgs) + + +@pytest.mark.parametrize('pkgs, expect_higher', [ + ( + [], + [], + ), + ( + [Package('spam', '3.2.1.9')], + [], # more precise but not strictly higher + ), + ( + [Package('spam', '3.3')], + ['spam-3.3'], # lower precision, but higher + ), + ( + [Package('spam', '3.2.1'), Package('eggs', '3.3.2')], + ['eggs-3.3.2'], # one too high + ), + ( + [Package('eggs', '1.2.3'), Package('eggs', '3.2.1.5'), Package('eggs', '3.4')], + ['eggs-3.4'], # multiple versions, one is higher + ), + ( + [Package('eggs', '3.2.1'), Package('eggs', '3.4'), Package('eggs', '3.3')], + ['eggs-3.4'], # multiple versions, two are higher + ), +]) +def test_check_pkgs_for_greater_version(pkgs, expect_higher): + if expect_higher: + with pytest.raises(aos_version.FoundHigherVersion) as e: + aos_version._check_higher_version_found(pkgs, expected_pkgs) + assert set(expect_higher) == set(e.value.problem_pkgs) + else: + aos_version._check_higher_version_found(pkgs, expected_pkgs) + + +@pytest.mark.parametrize('pkgs, expect_to_flag_pkgs', [ + ( + [], + [], + ), + ( + [Package('spam', '3.2.1')], + [], + ), + ( + [Package('spam', '3.2.1'), Package('eggs', '3.2.2')], + [], + ), + ( + [Package('spam', '3.2.1'), Package('spam', '3.3.2')], + ['spam'], + ), + ( + [Package('eggs', '1.2.3'), Package('eggs', '3.2.1.5'), Package('eggs', '3.4')], + ['eggs'], + ), +]) +def test_check_pkgs_for_multi_release(pkgs, expect_to_flag_pkgs): + if expect_to_flag_pkgs: + with pytest.raises(aos_version.FoundMultiRelease) as e: + aos_version._check_multi_minor_release(pkgs, expected_pkgs) + assert set(expect_to_flag_pkgs) == set(e.value.problem_pkgs) + else: + aos_version._check_multi_minor_release(pkgs, expected_pkgs) diff --git a/roles/openshift_health_checker/test/conftest.py b/roles/openshift_health_checker/test/conftest.py index bf717ae85..3cbd65507 100644 --- a/roles/openshift_health_checker/test/conftest.py +++ b/roles/openshift_health_checker/test/conftest.py @@ -1,5 +1,11 @@ import os import sys -# extend sys.path so that tests can import openshift_checks -sys.path.insert(1, os.path.dirname(os.path.dirname(__file__))) +# extend sys.path so that tests can import openshift_checks and action plugins +# from this role. +openshift_health_checker_path = os.path.dirname(os.path.dirname(__file__)) +sys.path[1:1] = [ + openshift_health_checker_path, + os.path.join(openshift_health_checker_path, 'action_plugins'), + os.path.join(openshift_health_checker_path, 'library'), +] diff --git a/roles/openshift_health_checker/test/curator_test.py b/roles/openshift_health_checker/test/curator_test.py new file mode 100644 index 000000000..ae108c96e --- /dev/null +++ b/roles/openshift_health_checker/test/curator_test.py @@ -0,0 +1,68 @@ +import pytest + +from openshift_checks.logging.curator import Curator + + +def canned_curator(exec_oc=None): + """Create a Curator check object with canned exec_oc method""" + check = Curator("dummy") # fails if a module is actually invoked + if exec_oc: + check._exec_oc = exec_oc + return check + + +def assert_error(error, expect_error): + if expect_error: + assert error + assert expect_error in error + else: + assert not error + + +plain_curator_pod = { + "metadata": { + "labels": {"component": "curator", "deploymentconfig": "logging-curator"}, + "name": "logging-curator-1", + }, + "status": { + "containerStatuses": [{"ready": True}], + "conditions": [{"status": "True", "type": "Ready"}], + "podIP": "10.10.10.10", + } +} + +not_running_curator_pod = { + "metadata": { + "labels": {"component": "curator", "deploymentconfig": "logging-curator"}, + "name": "logging-curator-2", + }, + "status": { + "containerStatuses": [{"ready": False}], + "conditions": [{"status": "False", "type": "Ready"}], + "podIP": "10.10.10.10", + } +} + + +@pytest.mark.parametrize('pods, expect_error', [ + ( + [], + "no Curator pods", + ), + ( + [plain_curator_pod], + None, + ), + ( + [not_running_curator_pod], + "not currently in a running state", + ), + ( + [plain_curator_pod, plain_curator_pod], + "more than one Curator pod", + ), +]) +def test_get_curator_pods(pods, expect_error): + check = canned_curator() + error = check.check_curator(pods) + assert_error(error, expect_error) diff --git a/roles/openshift_health_checker/test/disk_availability_test.py b/roles/openshift_health_checker/test/disk_availability_test.py new file mode 100644 index 000000000..945b9eafc --- /dev/null +++ b/roles/openshift_health_checker/test/disk_availability_test.py @@ -0,0 +1,180 @@ +import pytest + +from openshift_checks.disk_availability import DiskAvailability, OpenShiftCheckException + + +@pytest.mark.parametrize('group_names,is_active', [ + (['masters'], True), + (['nodes'], True), + (['etcd'], True), + (['masters', 'nodes'], True), + (['masters', 'etcd'], True), + ([], False), + (['lb'], False), + (['nfs'], False), +]) +def test_is_active(group_names, is_active): + task_vars = dict( + group_names=group_names, + ) + assert DiskAvailability.is_active(task_vars=task_vars) == is_active + + +@pytest.mark.parametrize('ansible_mounts,extra_words', [ + ([], ['none']), # empty ansible_mounts + ([{'mount': '/mnt'}], ['/mnt']), # missing relevant mount paths + ([{'mount': '/var'}], ['/var']), # missing size_available +]) +def test_cannot_determine_available_disk(ansible_mounts, extra_words): + task_vars = dict( + group_names=['masters'], + ansible_mounts=ansible_mounts, + ) + check = DiskAvailability(execute_module=fake_execute_module) + + with pytest.raises(OpenShiftCheckException) as excinfo: + check.run(tmp=None, task_vars=task_vars) + + for word in 'determine disk availability'.split() + extra_words: + assert word in str(excinfo.value) + + +@pytest.mark.parametrize('group_names,configured_min,ansible_mounts', [ + ( + ['masters'], + 0, + [{ + 'mount': '/', + 'size_available': 40 * 10**9 + 1, + }], + ), + ( + ['nodes'], + 0, + [{ + 'mount': '/', + 'size_available': 15 * 10**9 + 1, + }], + ), + ( + ['etcd'], + 0, + [{ + 'mount': '/', + 'size_available': 20 * 10**9 + 1, + }], + ), + ( + ['etcd'], + 1, # configure lower threshold + [{ + 'mount': '/', + 'size_available': 1 * 10**9 + 1, # way smaller than recommended + }], + ), + ( + ['etcd'], + 0, + [{ + # not enough space on / ... + 'mount': '/', + 'size_available': 2 * 10**9, + }, { + # ... but enough on /var + 'mount': '/var', + 'size_available': 20 * 10**9 + 1, + }], + ), +]) +def test_succeeds_with_recommended_disk_space(group_names, configured_min, ansible_mounts): + task_vars = dict( + group_names=group_names, + openshift_check_min_host_disk_gb=configured_min, + ansible_mounts=ansible_mounts, + ) + + check = DiskAvailability(execute_module=fake_execute_module) + result = check.run(tmp=None, task_vars=task_vars) + + assert not result.get('failed', False) + + +@pytest.mark.parametrize('group_names,configured_min,ansible_mounts,extra_words', [ + ( + ['masters'], + 0, + [{ + 'mount': '/', + 'size_available': 1, + }], + ['0.0 GB'], + ), + ( + ['masters'], + 100, # set a higher threshold + [{ + 'mount': '/', + 'size_available': 50 * 10**9, # would normally be enough... + }], + ['100.0 GB'], + ), + ( + ['nodes'], + 0, + [{ + 'mount': '/', + 'size_available': 1 * 10**9, + }], + ['1.0 GB'], + ), + ( + ['etcd'], + 0, + [{ + 'mount': '/', + 'size_available': 1, + }], + ['0.0 GB'], + ), + ( + ['nodes', 'masters'], + 0, + [{ + 'mount': '/', + # enough space for a node, not enough for a master + 'size_available': 15 * 10**9 + 1, + }], + ['15.0 GB'], + ), + ( + ['etcd'], + 0, + [{ + # enough space on / ... + 'mount': '/', + 'size_available': 20 * 10**9 + 1, + }, { + # .. but not enough on /var + 'mount': '/var', + 'size_available': 0, + }], + ['0.0 GB'], + ), +]) +def test_fails_with_insufficient_disk_space(group_names, configured_min, ansible_mounts, extra_words): + task_vars = dict( + group_names=group_names, + openshift_check_min_host_disk_gb=configured_min, + ansible_mounts=ansible_mounts, + ) + + check = DiskAvailability(execute_module=fake_execute_module) + result = check.run(tmp=None, task_vars=task_vars) + + assert result['failed'] + for word in 'below recommended'.split() + extra_words: + assert word in result['msg'] + + +def fake_execute_module(*args): + raise AssertionError('this function should not be called') diff --git a/roles/openshift_health_checker/test/docker_image_availability_test.py b/roles/openshift_health_checker/test/docker_image_availability_test.py index 2a9c32f77..0a7c0f8d3 100644 --- a/roles/openshift_health_checker/test/docker_image_availability_test.py +++ b/roles/openshift_health_checker/test/docker_image_availability_test.py @@ -3,26 +3,259 @@ import pytest from openshift_checks.docker_image_availability import DockerImageAvailability -@pytest.mark.xfail(strict=True) # TODO: remove this once this test is fully implemented. -@pytest.mark.parametrize('task_vars,expected_result', [ - ( - dict( - openshift=dict(common=dict( +@pytest.mark.parametrize('deployment_type, is_containerized, group_names, expect_active', [ + ("origin", True, [], True), + ("openshift-enterprise", True, [], True), + ("enterprise", True, [], False), + ("online", True, [], False), + ("invalid", True, [], False), + ("", True, [], False), + ("origin", False, [], False), + ("openshift-enterprise", False, [], False), + ("origin", False, ["nodes", "masters"], True), + ("openshift-enterprise", False, ["etcd"], False), +]) +def test_is_active(deployment_type, is_containerized, group_names, expect_active): + task_vars = dict( + openshift=dict(common=dict(is_containerized=is_containerized)), + openshift_deployment_type=deployment_type, + group_names=group_names, + ) + assert DockerImageAvailability.is_active(task_vars=task_vars) == expect_active + + +@pytest.mark.parametrize("is_containerized,is_atomic", [ + (True, True), + (False, False), + (True, False), + (False, True), +]) +def test_all_images_available_locally(is_containerized, is_atomic): + def execute_module(module_name, module_args, task_vars): + if module_name == "yum": + return {"changed": True} + + assert module_name == "docker_image_facts" + assert 'name' in module_args + assert module_args['name'] + return { + 'images': [module_args['name']], + } + + result = DockerImageAvailability(execute_module=execute_module).run(tmp=None, task_vars=dict( + openshift=dict( + common=dict( + service_type='origin', + is_containerized=is_containerized, + is_atomic=is_atomic, + ), + docker=dict(additional_registries=["docker.io"]), + ), + openshift_deployment_type='origin', + openshift_image_tag='3.4', + group_names=['nodes', 'masters'], + )) + + assert not result.get('failed', False) + + +@pytest.mark.parametrize("available_locally", [ + False, + True, +]) +def test_all_images_available_remotely(available_locally): + def execute_module(module_name, module_args, task_vars): + if module_name == 'docker_image_facts': + return {'images': [], 'failed': available_locally} + return {'changed': False} + + result = DockerImageAvailability(execute_module=execute_module).run(tmp=None, task_vars=dict( + openshift=dict( + common=dict( service_type='origin', is_containerized=False, - )), - openshift_release='v3.5', - deployment_type='origin', - openshift_image_tag='', # FIXME: should not be required + is_atomic=False, + ), + docker=dict(additional_registries=["docker.io", "registry.access.redhat.com"]), ), - {'changed': False}, + openshift_deployment_type='origin', + openshift_image_tag='v3.4', + group_names=['nodes', 'masters'], + )) + + assert not result.get('failed', False) + + +def test_all_images_unavailable(): + def execute_module(module_name=None, module_args=None, tmp=None, task_vars=None): + if module_name == "command": + return { + 'failed': True, + } + + return { + 'changed': False, + } + + check = DockerImageAvailability(execute_module=execute_module) + actual = check.run(tmp=None, task_vars=dict( + openshift=dict( + common=dict( + service_type='origin', + is_containerized=False, + is_atomic=False, + ), + docker=dict(additional_registries=["docker.io"]), + ), + openshift_deployment_type="openshift-enterprise", + openshift_image_tag='latest', + group_names=['nodes', 'masters'], + )) + + assert actual['failed'] + assert "required Docker images are not available" in actual['msg'] + + +@pytest.mark.parametrize("message,extra_words", [ + ( + "docker image update failure", + ["docker image update failure"], + ), + ( + "No package matching 'skopeo' found available, installed or updated", + ["dependencies can be installed via `yum`"] ), - # TODO: add more parameters here to test the multiple possible inputs that affect behavior. ]) -def test_docker_image_availability(task_vars, expected_result): +def test_skopeo_update_failure(message, extra_words): def execute_module(module_name=None, module_args=None, tmp=None, task_vars=None): - return {'info': {}} # TODO: this will vary depending on input parameters. + if module_name == "yum": + return { + "failed": True, + "msg": message, + "changed": False, + } - check = DockerImageAvailability(execute_module=execute_module) - result = check.run(tmp=None, task_vars=task_vars) - assert result == expected_result + return {'changed': False} + + actual = DockerImageAvailability(execute_module=execute_module).run(tmp=None, task_vars=dict( + openshift=dict( + common=dict( + service_type='origin', + is_containerized=False, + is_atomic=False, + ), + docker=dict(additional_registries=["unknown.io"]), + ), + openshift_deployment_type="openshift-enterprise", + openshift_image_tag='', + group_names=['nodes', 'masters'], + )) + + assert actual["failed"] + for word in extra_words: + assert word in actual["msg"] + + +@pytest.mark.parametrize("deployment_type,registries", [ + ("origin", ["unknown.io"]), + ("openshift-enterprise", ["registry.access.redhat.com"]), + ("openshift-enterprise", []), +]) +def test_registry_availability(deployment_type, registries): + def execute_module(module_name=None, module_args=None, tmp=None, task_vars=None): + return { + 'changed': False, + } + + actual = DockerImageAvailability(execute_module=execute_module).run(tmp=None, task_vars=dict( + openshift=dict( + common=dict( + service_type='origin', + is_containerized=False, + is_atomic=False, + ), + docker=dict(additional_registries=registries), + ), + openshift_deployment_type=deployment_type, + openshift_image_tag='', + group_names=['nodes', 'masters'], + )) + + assert not actual.get("failed", False) + + +@pytest.mark.parametrize("deployment_type, is_containerized, groups, oreg_url, expected", [ + ( # standard set of stuff required on nodes + "origin", False, ['nodes'], None, + set([ + 'openshift/origin-pod:vtest', + 'openshift/origin-deployer:vtest', + 'openshift/origin-docker-registry:vtest', + 'openshift/origin-haproxy-router:vtest', + 'cockpit/kubernetes', # origin version of registry-console + ]) + ), + ( # set a different URL for images + "origin", False, ['nodes'], 'foo.io/openshift/origin-${component}:${version}', + set([ + 'foo.io/openshift/origin-pod:vtest', + 'foo.io/openshift/origin-deployer:vtest', + 'foo.io/openshift/origin-docker-registry:vtest', + 'foo.io/openshift/origin-haproxy-router:vtest', + 'cockpit/kubernetes', # AFAICS this is not built from the URL + ]) + ), + ( + "origin", True, ['nodes', 'masters', 'etcd'], None, + set([ + # images running on top of openshift + 'openshift/origin-pod:vtest', + 'openshift/origin-deployer:vtest', + 'openshift/origin-docker-registry:vtest', + 'openshift/origin-haproxy-router:vtest', + 'cockpit/kubernetes', + # containerized component images + 'openshift/origin:vtest', + 'openshift/node:vtest', + 'openshift/openvswitch:vtest', + 'registry.access.redhat.com/rhel7/etcd', + ]) + ), + ( # enterprise images + "openshift-enterprise", True, ['nodes'], 'foo.io/openshift3/ose-${component}:f13ac45', + set([ + 'foo.io/openshift3/ose-pod:f13ac45', + 'foo.io/openshift3/ose-deployer:f13ac45', + 'foo.io/openshift3/ose-docker-registry:f13ac45', + 'foo.io/openshift3/ose-haproxy-router:f13ac45', + # registry-console is not constructed/versioned the same as the others. + 'registry.access.redhat.com/openshift3/registry-console', + # containerized images aren't built from oreg_url + 'openshift3/node:vtest', + 'openshift3/openvswitch:vtest', + ]) + ), + ( + "openshift-enterprise", True, ['etcd', 'lb'], 'foo.io/openshift3/ose-${component}:f13ac45', + set([ + 'registry.access.redhat.com/rhel7/etcd', + # lb does not yet come in a containerized version + ]) + ), + +]) +def test_required_images(deployment_type, is_containerized, groups, oreg_url, expected): + task_vars = dict( + openshift=dict( + common=dict( + is_containerized=is_containerized, + is_atomic=False, + ), + ), + openshift_deployment_type=deployment_type, + group_names=groups, + oreg_url=oreg_url, + openshift_image_tag='vtest', + ) + + assert expected == DockerImageAvailability("DUMMY").required_images(task_vars) diff --git a/roles/openshift_health_checker/test/docker_storage_test.py b/roles/openshift_health_checker/test/docker_storage_test.py new file mode 100644 index 000000000..876614b1d --- /dev/null +++ b/roles/openshift_health_checker/test/docker_storage_test.py @@ -0,0 +1,224 @@ +import pytest + +from openshift_checks import OpenShiftCheckException +from openshift_checks.docker_storage import DockerStorage + + +def dummy_check(execute_module=None): + def dummy_exec(self, status, task_vars): + raise Exception("dummy executor called") + return DockerStorage(execute_module=execute_module or dummy_exec) + + +@pytest.mark.parametrize('is_containerized, group_names, is_active', [ + (False, ["masters", "etcd"], False), + (False, ["masters", "nodes"], True), + (True, ["etcd"], True), +]) +def test_is_active(is_containerized, group_names, is_active): + task_vars = dict( + openshift=dict(common=dict(is_containerized=is_containerized)), + group_names=group_names, + ) + assert DockerStorage.is_active(task_vars=task_vars) == is_active + + +non_atomic_task_vars = {"openshift": {"common": {"is_atomic": False}}} + + +@pytest.mark.parametrize('docker_info, failed, expect_msg', [ + ( + dict(failed=True, msg="Error connecting: Error while fetching server API version"), + True, + ["Is docker running on this host?"], + ), + ( + dict(msg="I have no info"), + True, + ["missing info"], + ), + ( + dict(info={ + "Driver": "devicemapper", + "DriverStatus": [("Pool Name", "docker-docker--pool")], + }), + False, + [], + ), + ( + dict(info={ + "Driver": "devicemapper", + "DriverStatus": [("Data loop file", "true")], + }), + True, + ["loopback devices with the Docker devicemapper storage driver"], + ), + ( + dict(info={ + "Driver": "overlay2", + "DriverStatus": [] + }), + False, + [], + ), + ( + dict(info={ + "Driver": "overlay", + }), + True, + ["unsupported Docker storage driver"], + ), + ( + dict(info={ + "Driver": "unsupported", + }), + True, + ["unsupported Docker storage driver"], + ), +]) +def test_check_storage_driver(docker_info, failed, expect_msg): + def execute_module(module_name, module_args, tmp=None, task_vars=None): + if module_name == "yum": + return {} + if module_name != "docker_info": + raise ValueError("not expecting module " + module_name) + return docker_info + + check = dummy_check(execute_module=execute_module) + check._check_dm_usage = lambda status, task_vars: dict() # stub out for this test + result = check.run(tmp=None, task_vars=non_atomic_task_vars) + + if failed: + assert result["failed"] + else: + assert not result.get("failed", False) + + for word in expect_msg: + assert word in result["msg"] + + +enough_space = { + "Pool Name": "docker--vg-docker--pool", + "Data Space Used": "19.92 MB", + "Data Space Total": "8.535 GB", + "Metadata Space Used": "40.96 kB", + "Metadata Space Total": "25.17 MB", +} + +not_enough_space = { + "Pool Name": "docker--vg-docker--pool", + "Data Space Used": "10 GB", + "Data Space Total": "10 GB", + "Metadata Space Used": "42 kB", + "Metadata Space Total": "43 kB", +} + + +@pytest.mark.parametrize('task_vars, driver_status, vg_free, success, expect_msg', [ + ( + {"max_thinpool_data_usage_percent": "not a float"}, + enough_space, + "12g", + False, + ["is not a percentage"], + ), + ( + {}, + {}, # empty values from driver status + "bogus", # also does not parse as bytes + False, + ["Could not interpret", "as bytes"], + ), + ( + {}, + enough_space, + "12.00g", + True, + [], + ), + ( + {}, + not_enough_space, + "0.00", + False, + ["data usage", "metadata usage", "higher than threshold"], + ), +]) +def test_dm_usage(task_vars, driver_status, vg_free, success, expect_msg): + check = dummy_check() + check._get_vg_free = lambda pool, task_vars: vg_free + result = check._check_dm_usage(driver_status, task_vars) + result_success = not result.get("failed") + + assert result_success is success + for msg in expect_msg: + assert msg in result["msg"] + + +@pytest.mark.parametrize('pool, command_returns, raises, returns', [ + ( + "foo-bar", + { # vgs missing + "msg": "[Errno 2] No such file or directory", + "failed": True, + "cmd": "/sbin/vgs", + "rc": 2, + }, + "Failed to run /sbin/vgs", + None, + ), + ( + "foo", # no hyphen in name - should not happen + {}, + "name does not have the expected format", + None, + ), + ( + "foo-bar", + dict(stdout=" 4.00g\n"), + None, + "4.00g", + ), + ( + "foo-bar", + dict(stdout="\n"), # no matching VG + "vgs did not find this VG", + None, + ) +]) +def test_vg_free(pool, command_returns, raises, returns): + def execute_module(module_name, module_args, tmp=None, task_vars=None): + if module_name != "command": + raise ValueError("not expecting module " + module_name) + return command_returns + + check = dummy_check(execute_module=execute_module) + if raises: + with pytest.raises(OpenShiftCheckException) as err: + check._get_vg_free(pool, {}) + assert raises in str(err.value) + else: + ret = check._get_vg_free(pool, {}) + assert ret == returns + + +@pytest.mark.parametrize('string, expect_bytes', [ + ("12", 12.0), + ("12 k", 12.0 * 1024), + ("42.42 MB", 42.42 * 1024**2), + ("12g", 12.0 * 1024**3), +]) +def test_convert_to_bytes(string, expect_bytes): + got = DockerStorage._convert_to_bytes(string) + assert got == expect_bytes + + +@pytest.mark.parametrize('string', [ + "bork", + "42 Qs", +]) +def test_convert_to_bytes_error(string): + with pytest.raises(ValueError) as err: + DockerStorage._convert_to_bytes(string) + assert "Cannot convert" in str(err.value) + assert string in str(err.value) diff --git a/roles/openshift_health_checker/test/elasticsearch_test.py b/roles/openshift_health_checker/test/elasticsearch_test.py new file mode 100644 index 000000000..b9d375d8c --- /dev/null +++ b/roles/openshift_health_checker/test/elasticsearch_test.py @@ -0,0 +1,180 @@ +import pytest +import json + +from openshift_checks.logging.elasticsearch import Elasticsearch + +task_vars_config_base = dict(openshift=dict(common=dict(config_base='/etc/origin'))) + + +def canned_elasticsearch(exec_oc=None): + """Create an Elasticsearch check object with canned exec_oc method""" + check = Elasticsearch("dummy") # fails if a module is actually invoked + if exec_oc: + check._exec_oc = exec_oc + return check + + +def assert_error(error, expect_error): + if expect_error: + assert error + assert expect_error in error + else: + assert not error + + +plain_es_pod = { + "metadata": { + "labels": {"component": "es", "deploymentconfig": "logging-es"}, + "name": "logging-es", + }, + "status": { + "conditions": [{"status": "True", "type": "Ready"}], + "containerStatuses": [{"ready": True}], + "podIP": "10.10.10.10", + }, + "_test_master_name_str": "name logging-es", +} + +split_es_pod = { + "metadata": { + "labels": {"component": "es", "deploymentconfig": "logging-es-2"}, + "name": "logging-es-2", + }, + "status": { + "conditions": [{"status": "True", "type": "Ready"}], + "containerStatuses": [{"ready": True}], + "podIP": "10.10.10.10", + }, + "_test_master_name_str": "name logging-es-2", +} + + +def test_check_elasticsearch(): + assert 'No logging Elasticsearch pods' in canned_elasticsearch().check_elasticsearch([], {}) + + # canned oc responses to match so all the checks pass + def _exec_oc(cmd, args, task_vars): + if '_cat/master' in cmd: + return 'name logging-es' + elif '/_nodes' in cmd: + return json.dumps(es_node_list) + elif '_cluster/health' in cmd: + return '{"status": "green"}' + elif ' df ' in cmd: + return 'IUse% Use%\n 3% 4%\n' + else: + raise Exception(cmd) + + assert not canned_elasticsearch(_exec_oc).check_elasticsearch([plain_es_pod], {}) + + +def pods_by_name(pods): + return {pod['metadata']['name']: pod for pod in pods} + + +@pytest.mark.parametrize('pods, expect_error', [ + ( + [], + 'No logging Elasticsearch masters', + ), + ( + [plain_es_pod], + None, + ), + ( + [plain_es_pod, split_es_pod], + 'Found multiple Elasticsearch masters', + ), +]) +def test_check_elasticsearch_masters(pods, expect_error): + test_pods = list(pods) + check = canned_elasticsearch(lambda cmd, args, task_vars: test_pods.pop(0)['_test_master_name_str']) + + errors = check._check_elasticsearch_masters(pods_by_name(pods), task_vars_config_base) + assert_error(''.join(errors), expect_error) + + +es_node_list = { + 'nodes': { + 'random-es-name': { + 'host': 'logging-es', + }}} + + +@pytest.mark.parametrize('pods, node_list, expect_error', [ + ( + [], + {}, + 'No logging Elasticsearch masters', + ), + ( + [plain_es_pod], + es_node_list, + None, + ), + ( + [plain_es_pod], + {}, # empty list of nodes triggers KeyError + "Failed to query", + ), + ( + [split_es_pod], + es_node_list, + 'does not correspond to any known ES pod', + ), +]) +def test_check_elasticsearch_node_list(pods, node_list, expect_error): + check = canned_elasticsearch(lambda cmd, args, task_vars: json.dumps(node_list)) + + errors = check._check_elasticsearch_node_list(pods_by_name(pods), task_vars_config_base) + assert_error(''.join(errors), expect_error) + + +@pytest.mark.parametrize('pods, health_data, expect_error', [ + ( + [plain_es_pod], + [{"status": "green"}], + None, + ), + ( + [plain_es_pod], + [{"no-status": "should bomb"}], + 'Could not retrieve cluster health status', + ), + ( + [plain_es_pod, split_es_pod], + [{"status": "green"}, {"status": "red"}], + 'Elasticsearch cluster health status is RED', + ), +]) +def test_check_elasticsearch_cluster_health(pods, health_data, expect_error): + test_health_data = list(health_data) + check = canned_elasticsearch(lambda cmd, args, task_vars: json.dumps(test_health_data.pop(0))) + + errors = check._check_es_cluster_health(pods_by_name(pods), task_vars_config_base) + assert_error(''.join(errors), expect_error) + + +@pytest.mark.parametrize('disk_data, expect_error', [ + ( + 'df: /elasticsearch/persistent: No such file or directory\n', + 'Could not retrieve storage usage', + ), + ( + 'IUse% Use%\n 3% 4%\n', + None, + ), + ( + 'IUse% Use%\n 95% 40%\n', + 'Inode percent usage on the storage volume', + ), + ( + 'IUse% Use%\n 3% 94%\n', + 'Disk percent usage on the storage volume', + ), +]) +def test_check_elasticsearch_diskspace(disk_data, expect_error): + check = canned_elasticsearch(lambda cmd, args, task_vars: disk_data) + + errors = check._check_elasticsearch_diskspace(pods_by_name([plain_es_pod]), task_vars_config_base) + assert_error(''.join(errors), expect_error) diff --git a/roles/openshift_health_checker/test/etcd_imagedata_size_test.py b/roles/openshift_health_checker/test/etcd_imagedata_size_test.py new file mode 100644 index 000000000..df9d52d41 --- /dev/null +++ b/roles/openshift_health_checker/test/etcd_imagedata_size_test.py @@ -0,0 +1,328 @@ +import pytest + +from collections import namedtuple +from openshift_checks.etcd_imagedata_size import EtcdImageDataSize, OpenShiftCheckException +from etcdkeysize import check_etcd_key_size + + +def fake_etcd_client(root): + fake_nodes = dict() + fake_etcd_node(root, fake_nodes) + + clientclass = namedtuple("client", ["read"]) + return clientclass(lambda key, recursive: fake_etcd_result(fake_nodes[key])) + + +def fake_etcd_result(fake_node): + resultclass = namedtuple("result", ["leaves"]) + if not fake_node.dir: + return resultclass([fake_node]) + + return resultclass(fake_node.leaves) + + +def fake_etcd_node(node, visited): + min_req_fields = ["dir", "key"] + fields = list(node) + leaves = [] + + if node["dir"] and node.get("leaves"): + for leaf in node["leaves"]: + leaves.append(fake_etcd_node(leaf, visited)) + + if len(set(min_req_fields) - set(fields)) > 0: + raise ValueError("fake etcd nodes require at least {} fields.".format(min_req_fields)) + + if node.get("leaves"): + node["leaves"] = leaves + + nodeclass = namedtuple("node", fields) + nodeinst = nodeclass(**node) + visited[nodeinst.key] = nodeinst + + return nodeinst + + +@pytest.mark.parametrize('ansible_mounts,extra_words', [ + ([], ['none']), # empty ansible_mounts + ([{'mount': '/mnt'}], ['/mnt']), # missing relevant mount paths +]) +def test_cannot_determine_available_mountpath(ansible_mounts, extra_words): + task_vars = dict( + ansible_mounts=ansible_mounts, + ) + check = EtcdImageDataSize(execute_module=fake_execute_module) + + with pytest.raises(OpenShiftCheckException) as excinfo: + check.run(tmp=None, task_vars=task_vars) + + for word in 'determine valid etcd mountpath'.split() + extra_words: + assert word in str(excinfo.value) + + +@pytest.mark.parametrize('ansible_mounts,tree,size_limit,should_fail,extra_words', [ + ( + # test that default image size limit evals to 1/2 * (total size in use) + [{ + 'mount': '/', + 'size_available': 40 * 10**9, + 'size_total': 80 * 10**9, + }], + {"dir": False, "key": "/", "value": "1234"}, + None, + False, + [], + ), + ( + [{ + 'mount': '/', + 'size_available': 40 * 10**9, + 'size_total': 48 * 10**9, + }], + {"dir": False, "key": "/", "value": "1234"}, + None, + False, + [], + ), + ( + # set max size limit for image data to be below total node value + # total node value is defined as the sum of the value field + # from every node + [{ + 'mount': '/', + 'size_available': 40 * 10**9, + 'size_total': 48 * 10**9, + }], + {"dir": False, "key": "/", "value": "12345678"}, + 7, + True, + ["exceeds the maximum recommended limit", "0.00 GB"], + ), + ( + [{ + 'mount': '/', + 'size_available': 48 * 10**9 - 1, + 'size_total': 48 * 10**9, + }], + {"dir": False, "key": "/", "value": "1234"}, + None, + True, + ["exceeds the maximum recommended limit", "0.00 GB"], + ) +]) +def test_check_etcd_key_size_calculates_correct_limit(ansible_mounts, tree, size_limit, should_fail, extra_words): + def execute_module(module_name, args, tmp=None, task_vars=None): + if module_name != "etcdkeysize": + return { + "changed": False, + } + + client = fake_etcd_client(tree) + s, limit_exceeded = check_etcd_key_size(client, tree["key"], args["size_limit_bytes"]) + + return {"size_limit_exceeded": limit_exceeded} + + task_vars = dict( + etcd_max_image_data_size_bytes=size_limit, + ansible_mounts=ansible_mounts, + openshift=dict( + master=dict(etcd_hosts=["localhost"]), + common=dict(config_base="/var/lib/origin") + ) + ) + if size_limit is None: + task_vars.pop("etcd_max_image_data_size_bytes") + + check = EtcdImageDataSize(execute_module=execute_module).run(tmp=None, task_vars=task_vars) + + if should_fail: + assert check["failed"] + + for word in extra_words: + assert word in check["msg"] + else: + assert not check.get("failed", False) + + +@pytest.mark.parametrize('ansible_mounts,tree,root_path,expected_size,extra_words', [ + ( + [{ + 'mount': '/', + 'size_available': 40 * 10**9, + 'size_total': 80 * 10**9, + }], + # test recursive size check on tree with height > 1 + { + "dir": True, + "key": "/", + "leaves": [ + {"dir": False, "key": "/foo1", "value": "1234"}, + {"dir": False, "key": "/foo2", "value": "1234"}, + {"dir": False, "key": "/foo3", "value": "1234"}, + {"dir": False, "key": "/foo4", "value": "1234"}, + { + "dir": True, + "key": "/foo5", + "leaves": [ + {"dir": False, "key": "/foo/bar1", "value": "56789"}, + {"dir": False, "key": "/foo/bar2", "value": "56789"}, + {"dir": False, "key": "/foo/bar3", "value": "56789"}, + { + "dir": True, + "key": "/foo/bar4", + "leaves": [ + {"dir": False, "key": "/foo/bar/baz1", "value": "123"}, + {"dir": False, "key": "/foo/bar/baz2", "value": "123"}, + ] + }, + ] + }, + ] + }, + "/", + 37, + [], + ), + ( + [{ + 'mount': '/', + 'size_available': 40 * 10**9, + 'size_total': 80 * 10**9, + }], + # test correct sub-tree size calculation + { + "dir": True, + "key": "/", + "leaves": [ + {"dir": False, "key": "/foo1", "value": "1234"}, + {"dir": False, "key": "/foo2", "value": "1234"}, + {"dir": False, "key": "/foo3", "value": "1234"}, + {"dir": False, "key": "/foo4", "value": "1234"}, + { + "dir": True, + "key": "/foo5", + "leaves": [ + {"dir": False, "key": "/foo/bar1", "value": "56789"}, + {"dir": False, "key": "/foo/bar2", "value": "56789"}, + {"dir": False, "key": "/foo/bar3", "value": "56789"}, + { + "dir": True, + "key": "/foo/bar4", + "leaves": [ + {"dir": False, "key": "/foo/bar/baz1", "value": "123"}, + {"dir": False, "key": "/foo/bar/baz2", "value": "123"}, + ] + }, + ] + }, + ] + }, + "/foo5", + 21, + [], + ), + ( + [{ + 'mount': '/', + 'size_available': 40 * 10**9, + 'size_total': 80 * 10**9, + }], + # test that a non-existing key is handled correctly + { + "dir": False, + "key": "/", + "value": "1234", + }, + "/missing", + 0, + [], + ), + ( + [{ + 'mount': '/', + 'size_available': 40 * 10**9, + 'size_total': 80 * 10**9, + }], + # test etcd cycle handling + { + "dir": True, + "key": "/", + "leaves": [ + {"dir": False, "key": "/foo1", "value": "1234"}, + {"dir": False, "key": "/foo2", "value": "1234"}, + {"dir": False, "key": "/foo3", "value": "1234"}, + {"dir": False, "key": "/foo4", "value": "1234"}, + { + "dir": True, + "key": "/", + "leaves": [ + {"dir": False, "key": "/foo1", "value": "1"}, + ], + }, + ] + }, + "/", + 16, + [], + ), +]) +def test_etcd_key_size_check_calculates_correct_size(ansible_mounts, tree, root_path, expected_size, extra_words): + def execute_module(module_name, args, tmp=None, task_vars=None): + if module_name != "etcdkeysize": + return { + "changed": False, + } + + client = fake_etcd_client(tree) + size, limit_exceeded = check_etcd_key_size(client, root_path, args["size_limit_bytes"]) + + assert size == expected_size + return { + "size_limit_exceeded": limit_exceeded, + } + + task_vars = dict( + ansible_mounts=ansible_mounts, + openshift=dict( + master=dict(etcd_hosts=["localhost"]), + common=dict(config_base="/var/lib/origin") + ) + ) + + check = EtcdImageDataSize(execute_module=execute_module).run(tmp=None, task_vars=task_vars) + assert not check.get("failed", False) + + +def test_etcdkeysize_module_failure(): + def execute_module(module_name, tmp=None, task_vars=None): + if module_name != "etcdkeysize": + return { + "changed": False, + } + + return { + "rc": 1, + "module_stderr": "failure", + } + + task_vars = dict( + ansible_mounts=[{ + 'mount': '/', + 'size_available': 40 * 10**9, + 'size_total': 80 * 10**9, + }], + openshift=dict( + master=dict(etcd_hosts=["localhost"]), + common=dict(config_base="/var/lib/origin") + ) + ) + + check = EtcdImageDataSize(execute_module=execute_module).run(tmp=None, task_vars=task_vars) + + assert check["failed"] + for word in "Failed to retrieve stats": + assert word in check["msg"] + + +def fake_execute_module(*args): + raise AssertionError('this function should not be called') diff --git a/roles/openshift_health_checker/test/etcd_volume_test.py b/roles/openshift_health_checker/test/etcd_volume_test.py new file mode 100644 index 000000000..917045526 --- /dev/null +++ b/roles/openshift_health_checker/test/etcd_volume_test.py @@ -0,0 +1,149 @@ +import pytest + +from openshift_checks.etcd_volume import EtcdVolume, OpenShiftCheckException + + +@pytest.mark.parametrize('ansible_mounts,extra_words', [ + ([], ['none']), # empty ansible_mounts + ([{'mount': '/mnt'}], ['/mnt']), # missing relevant mount paths +]) +def test_cannot_determine_available_disk(ansible_mounts, extra_words): + task_vars = dict( + ansible_mounts=ansible_mounts, + ) + check = EtcdVolume(execute_module=fake_execute_module) + + with pytest.raises(OpenShiftCheckException) as excinfo: + check.run(tmp=None, task_vars=task_vars) + + for word in 'Unable to find etcd storage mount point'.split() + extra_words: + assert word in str(excinfo.value) + + +@pytest.mark.parametrize('size_limit,ansible_mounts', [ + ( + # if no size limit is specified, expect max usage + # limit to default to 90% of size_total + None, + [{ + 'mount': '/', + 'size_available': 40 * 10**9, + 'size_total': 80 * 10**9 + }], + ), + ( + 1, + [{ + 'mount': '/', + 'size_available': 30 * 10**9, + 'size_total': 30 * 10**9, + }], + ), + ( + 20000000000, + [{ + 'mount': '/', + 'size_available': 20 * 10**9, + 'size_total': 40 * 10**9, + }], + ), + ( + 5000000000, + [{ + # not enough space on / ... + 'mount': '/', + 'size_available': 0, + 'size_total': 0, + }, { + # not enough space on /var/lib ... + 'mount': '/var/lib', + 'size_available': 2 * 10**9, + 'size_total': 21 * 10**9, + }, { + # ... but enough on /var/lib/etcd + 'mount': '/var/lib/etcd', + 'size_available': 36 * 10**9, + 'size_total': 40 * 10**9 + }], + ) +]) +def test_succeeds_with_recommended_disk_space(size_limit, ansible_mounts): + task_vars = dict( + etcd_device_usage_threshold_percent=size_limit, + ansible_mounts=ansible_mounts, + ) + + if task_vars["etcd_device_usage_threshold_percent"] is None: + task_vars.pop("etcd_device_usage_threshold_percent") + + check = EtcdVolume(execute_module=fake_execute_module) + result = check.run(tmp=None, task_vars=task_vars) + + assert not result.get('failed', False) + + +@pytest.mark.parametrize('size_limit_percent,ansible_mounts,extra_words', [ + ( + # if no size limit is specified, expect max usage + # limit to default to 90% of size_total + None, + [{ + 'mount': '/', + 'size_available': 1 * 10**9, + 'size_total': 100 * 10**9, + }], + ['99.0%'], + ), + ( + 70.0, + [{ + 'mount': '/', + 'size_available': 1 * 10**6, + 'size_total': 5 * 10**9, + }], + ['100.0%'], + ), + ( + 40.0, + [{ + 'mount': '/', + 'size_available': 2 * 10**9, + 'size_total': 6 * 10**9, + }], + ['66.7%'], + ), + ( + None, + [{ + # enough space on /var ... + 'mount': '/var', + 'size_available': 20 * 10**9, + 'size_total': 20 * 10**9, + }, { + # .. but not enough on /var/lib + 'mount': '/var/lib', + 'size_available': 1 * 10**9, + 'size_total': 20 * 10**9, + }], + ['95.0%'], + ), +]) +def test_fails_with_insufficient_disk_space(size_limit_percent, ansible_mounts, extra_words): + task_vars = dict( + etcd_device_usage_threshold_percent=size_limit_percent, + ansible_mounts=ansible_mounts, + ) + + if task_vars["etcd_device_usage_threshold_percent"] is None: + task_vars.pop("etcd_device_usage_threshold_percent") + + check = EtcdVolume(execute_module=fake_execute_module) + result = check.run(tmp=None, task_vars=task_vars) + + assert result['failed'] + for word in extra_words: + assert word in result['msg'] + + +def fake_execute_module(*args): + raise AssertionError('this function should not be called') diff --git a/roles/openshift_health_checker/test/fluentd_test.py b/roles/openshift_health_checker/test/fluentd_test.py new file mode 100644 index 000000000..d151c0b19 --- /dev/null +++ b/roles/openshift_health_checker/test/fluentd_test.py @@ -0,0 +1,109 @@ +import pytest +import json + +from openshift_checks.logging.fluentd import Fluentd + + +def canned_fluentd(exec_oc=None): + """Create a Fluentd check object with canned exec_oc method""" + check = Fluentd("dummy") # fails if a module is actually invoked + if exec_oc: + check._exec_oc = exec_oc + return check + + +def assert_error(error, expect_error): + if expect_error: + assert error + assert expect_error in error + else: + assert not error + + +fluentd_pod_node1 = { + "metadata": { + "labels": {"component": "fluentd", "deploymentconfig": "logging-fluentd"}, + "name": "logging-fluentd-1", + }, + "spec": {"host": "node1", "nodeName": "node1"}, + "status": { + "containerStatuses": [{"ready": True}], + "conditions": [{"status": "True", "type": "Ready"}], + } +} +fluentd_pod_node2_down = { + "metadata": { + "labels": {"component": "fluentd", "deploymentconfig": "logging-fluentd"}, + "name": "logging-fluentd-2", + }, + "spec": {"host": "node2", "nodeName": "node2"}, + "status": { + "containerStatuses": [{"ready": False}], + "conditions": [{"status": "False", "type": "Ready"}], + } +} +fluentd_node1 = { + "metadata": { + "labels": {"logging-infra-fluentd": "true", "kubernetes.io/hostname": "node1"}, + "name": "node1", + }, + "status": {"addresses": [{"type": "InternalIP", "address": "10.10.1.1"}]}, +} +fluentd_node2 = { + "metadata": { + "labels": {"logging-infra-fluentd": "true", "kubernetes.io/hostname": "hostname"}, + "name": "node2", + }, + "status": {"addresses": [{"type": "InternalIP", "address": "10.10.1.2"}]}, +} +fluentd_node3_unlabeled = { + "metadata": { + "labels": {"kubernetes.io/hostname": "hostname"}, + "name": "node3", + }, + "status": {"addresses": [{"type": "InternalIP", "address": "10.10.1.3"}]}, +} + + +@pytest.mark.parametrize('pods, nodes, expect_error', [ + ( + [], + [], + 'No nodes appear to be defined', + ), + ( + [], + [fluentd_node3_unlabeled], + 'There are no nodes with the fluentd label', + ), + ( + [], + [fluentd_node1, fluentd_node3_unlabeled], + 'Fluentd will not aggregate logs from these nodes.', + ), + ( + [], + [fluentd_node2], + "nodes are supposed to have a Fluentd pod but do not", + ), + ( + [fluentd_pod_node1, fluentd_pod_node1], + [fluentd_node1], + 'more Fluentd pods running than nodes labeled', + ), + ( + [fluentd_pod_node2_down], + [fluentd_node2], + "Fluentd pods are supposed to be running", + ), + ( + [fluentd_pod_node1], + [fluentd_node1], + None, + ), +]) +def test_get_fluentd_pods(pods, nodes, expect_error): + check = canned_fluentd(lambda cmd, args, task_vars: json.dumps(dict(items=nodes))) + + error = check.check_fluentd(pods, {}) + assert_error(error, expect_error) diff --git a/roles/openshift_health_checker/test/kibana_test.py b/roles/openshift_health_checker/test/kibana_test.py new file mode 100644 index 000000000..19140a1b6 --- /dev/null +++ b/roles/openshift_health_checker/test/kibana_test.py @@ -0,0 +1,218 @@ +import pytest +import json + +try: + import urllib2 + from urllib2 import HTTPError, URLError +except ImportError: + from urllib.error import HTTPError, URLError + import urllib.request as urllib2 + +from openshift_checks.logging.kibana import Kibana + + +def canned_kibana(exec_oc=None): + """Create a Kibana check object with canned exec_oc method""" + check = Kibana("dummy") # fails if a module is actually invoked + if exec_oc: + check._exec_oc = exec_oc + return check + + +def assert_error(error, expect_error): + if expect_error: + assert error + assert expect_error in error + else: + assert not error + + +plain_kibana_pod = { + "metadata": { + "labels": {"component": "kibana", "deploymentconfig": "logging-kibana"}, + "name": "logging-kibana-1", + }, + "status": { + "containerStatuses": [{"ready": True}, {"ready": True}], + "conditions": [{"status": "True", "type": "Ready"}], + } +} +not_running_kibana_pod = { + "metadata": { + "labels": {"component": "kibana", "deploymentconfig": "logging-kibana"}, + "name": "logging-kibana-2", + }, + "status": { + "containerStatuses": [{"ready": True}, {"ready": False}], + "conditions": [{"status": "True", "type": "Ready"}], + } +} + + +@pytest.mark.parametrize('pods, expect_error', [ + ( + [], + "There are no Kibana pods deployed", + ), + ( + [plain_kibana_pod], + None, + ), + ( + [not_running_kibana_pod], + "No Kibana pod is in a running state", + ), + ( + [plain_kibana_pod, not_running_kibana_pod], + "The following Kibana pods are not currently in a running state", + ), +]) +def test_check_kibana(pods, expect_error): + check = canned_kibana() + error = check.check_kibana(pods) + assert_error(error, expect_error) + + +@pytest.mark.parametrize('route, expect_url, expect_error', [ + ( + None, + None, + 'no_route_exists', + ), + + # test route with no ingress + ( + { + "metadata": { + "labels": {"component": "kibana", "deploymentconfig": "logging-kibana"}, + "name": "logging-kibana", + }, + "status": { + "ingress": [], + }, + "spec": { + "host": "hostname", + } + }, + None, + 'route_not_accepted', + ), + + # test route with no host + ( + { + "metadata": { + "labels": {"component": "kibana", "deploymentconfig": "logging-kibana"}, + "name": "logging-kibana", + }, + "status": { + "ingress": [{ + "status": True, + }], + }, + "spec": {}, + }, + None, + 'route_missing_host', + ), + + # test route that looks fine + ( + { + "metadata": { + "labels": {"component": "kibana", "deploymentconfig": "logging-kibana"}, + "name": "logging-kibana", + }, + "status": { + "ingress": [{ + "status": True, + }], + }, + "spec": { + "host": "hostname", + }, + }, + "https://hostname/", + None, + ), +]) +def test_get_kibana_url(route, expect_url, expect_error): + check = canned_kibana(lambda cmd, args, task_vars: json.dumps(route) if route else "") + + url, error = check._get_kibana_url({}) + if expect_url: + assert url == expect_url + else: + assert not url + if expect_error: + assert error == expect_error + else: + assert not error + + +@pytest.mark.parametrize('exec_result, expect', [ + ( + 'urlopen error [Errno 111] Connection refused', + 'at least one router routing to it?', + ), + ( + 'urlopen error [Errno -2] Name or service not known', + 'DNS configured for the Kibana hostname?', + ), + ( + 'Status code was not [302]: HTTP Error 500: Server error', + 'did not return the correct status code', + ), + ( + 'bork bork bork', + 'bork bork bork', # should pass through + ), +]) +def test_verify_url_internal_failure(exec_result, expect): + check = Kibana(execute_module=lambda module_name, args, task_vars: dict(failed=True, msg=exec_result)) + check._get_kibana_url = lambda task_vars: ('url', None) + + error = check._check_kibana_route({}) + assert_error(error, expect) + + +@pytest.mark.parametrize('lib_result, expect', [ + ( + HTTPError('url', 500, "it broke", hdrs=None, fp=None), + 'it broke', + ), + ( + URLError('it broke'), + 'it broke', + ), + ( + 302, + 'returned the wrong error code', + ), + ( + 200, + None, + ), +]) +def test_verify_url_external_failure(lib_result, expect, monkeypatch): + + class _http_return: + + def __init__(self, code): + self.code = code + + def getcode(self): + return self.code + + def urlopen(url, context): + if type(lib_result) is int: + return _http_return(lib_result) + raise lib_result + monkeypatch.setattr(urllib2, 'urlopen', urlopen) + + check = canned_kibana() + check._get_kibana_url = lambda task_vars: ('url', None) + check._verify_url_internal = lambda url, task_vars: None + + error = check._check_kibana_route({}) + assert_error(error, expect) diff --git a/roles/openshift_health_checker/test/logging_check_test.py b/roles/openshift_health_checker/test/logging_check_test.py new file mode 100644 index 000000000..b6db34fe3 --- /dev/null +++ b/roles/openshift_health_checker/test/logging_check_test.py @@ -0,0 +1,137 @@ +import pytest +import json + +from openshift_checks.logging.logging import LoggingCheck, OpenShiftCheckException + +task_vars_config_base = dict(openshift=dict(common=dict(config_base='/etc/origin'))) + + +logging_namespace = "logging" + + +def canned_loggingcheck(exec_oc=None): + """Create a LoggingCheck object with canned exec_oc method""" + check = LoggingCheck("dummy") # fails if a module is actually invoked + check.logging_namespace = 'logging' + if exec_oc: + check.exec_oc = exec_oc + return check + + +def assert_error(error, expect_error): + if expect_error: + assert error + assert expect_error in error + else: + assert not error + + +plain_es_pod = { + "metadata": { + "labels": {"component": "es", "deploymentconfig": "logging-es"}, + "name": "logging-es", + }, + "status": { + "conditions": [{"status": "True", "type": "Ready"}], + "containerStatuses": [{"ready": True}], + "podIP": "10.10.10.10", + }, + "_test_master_name_str": "name logging-es", +} + +plain_kibana_pod = { + "metadata": { + "labels": {"component": "kibana", "deploymentconfig": "logging-kibana"}, + "name": "logging-kibana-1", + }, + "status": { + "containerStatuses": [{"ready": True}, {"ready": True}], + "conditions": [{"status": "True", "type": "Ready"}], + } +} + +fluentd_pod_node1 = { + "metadata": { + "labels": {"component": "fluentd", "deploymentconfig": "logging-fluentd"}, + "name": "logging-fluentd-1", + }, + "spec": {"host": "node1", "nodeName": "node1"}, + "status": { + "containerStatuses": [{"ready": True}], + "conditions": [{"status": "True", "type": "Ready"}], + } +} + +plain_curator_pod = { + "metadata": { + "labels": {"component": "curator", "deploymentconfig": "logging-curator"}, + "name": "logging-curator-1", + }, + "status": { + "containerStatuses": [{"ready": True}], + "conditions": [{"status": "True", "type": "Ready"}], + "podIP": "10.10.10.10", + } +} + + +@pytest.mark.parametrize('problem, expect', [ + ("[Errno 2] No such file or directory", "supposed to be a master"), + ("Permission denied", "Unexpected error using `oc`"), +]) +def test_oc_failure(problem, expect): + def execute_module(module_name, args, task_vars): + if module_name == "ocutil": + return dict(failed=True, result=problem) + return dict(changed=False) + + check = LoggingCheck({}) + + with pytest.raises(OpenShiftCheckException) as excinfo: + check.exec_oc(execute_module, logging_namespace, 'get foo', [], task_vars=task_vars_config_base) + assert expect in str(excinfo) + + +groups_with_first_master = dict(masters=['this-host', 'other-host']) +groups_with_second_master = dict(masters=['other-host', 'this-host']) +groups_not_a_master = dict(masters=['other-host']) + + +@pytest.mark.parametrize('groups, logging_deployed, is_active', [ + (groups_with_first_master, True, True), + (groups_with_first_master, False, False), + (groups_not_a_master, True, False), + (groups_with_second_master, True, False), + (groups_not_a_master, True, False), +]) +def test_is_active(groups, logging_deployed, is_active): + task_vars = dict( + ansible_ssh_host='this-host', + groups=groups, + openshift_hosted_logging_deploy=logging_deployed, + ) + + assert LoggingCheck.is_active(task_vars=task_vars) == is_active + + +@pytest.mark.parametrize('pod_output, expect_pods, expect_error', [ + ( + 'No resources found.', + None, + 'There are no pods in the logging namespace', + ), + ( + json.dumps({'items': [plain_kibana_pod, plain_es_pod, plain_curator_pod, fluentd_pod_node1]}), + [plain_es_pod], + None, + ), +]) +def test_get_pods_for_component(pod_output, expect_pods, expect_error): + check = canned_loggingcheck(lambda exec_module, namespace, cmd, args, task_vars: pod_output) + pods, error = check.get_pods_for_component( + lambda name, args, task_vars: {}, + logging_namespace, + "es", + {} + ) + assert_error(error, expect_error) diff --git a/roles/openshift_health_checker/test/memory_availability_test.py b/roles/openshift_health_checker/test/memory_availability_test.py new file mode 100644 index 000000000..4fbaea0a9 --- /dev/null +++ b/roles/openshift_health_checker/test/memory_availability_test.py @@ -0,0 +1,129 @@ +import pytest + +from openshift_checks.memory_availability import MemoryAvailability + + +@pytest.mark.parametrize('group_names,is_active', [ + (['masters'], True), + (['nodes'], True), + (['etcd'], True), + (['masters', 'nodes'], True), + (['masters', 'etcd'], True), + ([], False), + (['lb'], False), + (['nfs'], False), +]) +def test_is_active(group_names, is_active): + task_vars = dict( + group_names=group_names, + ) + assert MemoryAvailability.is_active(task_vars=task_vars) == is_active + + +@pytest.mark.parametrize('group_names,configured_min,ansible_memtotal_mb', [ + ( + ['masters'], + 0, + 17200, + ), + ( + ['nodes'], + 0, + 8200, + ), + ( + ['nodes'], + 1, # configure lower threshold + 2000, # too low for recommended but not for configured + ), + ( + ['nodes'], + 2, # configure threshold where adjustment pushes it over + 1900, + ), + ( + ['etcd'], + 0, + 8200, + ), + ( + ['masters', 'nodes'], + 0, + 17000, + ), +]) +def test_succeeds_with_recommended_memory(group_names, configured_min, ansible_memtotal_mb): + task_vars = dict( + group_names=group_names, + openshift_check_min_host_memory_gb=configured_min, + ansible_memtotal_mb=ansible_memtotal_mb, + ) + + check = MemoryAvailability(execute_module=fake_execute_module) + result = check.run(tmp=None, task_vars=task_vars) + + assert not result.get('failed', False) + + +@pytest.mark.parametrize('group_names,configured_min,ansible_memtotal_mb,extra_words', [ + ( + ['masters'], + 0, + 0, + ['0.0 GiB'], + ), + ( + ['nodes'], + 0, + 100, + ['0.1 GiB'], + ), + ( + ['nodes'], + 24, # configure higher threshold + 20 * 1024, # enough to meet recommended but not configured + ['20.0 GiB'], + ), + ( + ['nodes'], + 24, # configure higher threshold + 22 * 1024, # not enough for adjustment to push over threshold + ['22.0 GiB'], + ), + ( + ['etcd'], + 0, + 6 * 1024, + ['6.0 GiB'], + ), + ( + ['etcd', 'masters'], + 0, + 9 * 1024, # enough memory for etcd, not enough for a master + ['9.0 GiB'], + ), + ( + ['nodes', 'masters'], + 0, + # enough memory for a node, not enough for a master + 11 * 1024, + ['11.0 GiB'], + ), +]) +def test_fails_with_insufficient_memory(group_names, configured_min, ansible_memtotal_mb, extra_words): + task_vars = dict( + group_names=group_names, + openshift_check_min_host_memory_gb=configured_min, + ansible_memtotal_mb=ansible_memtotal_mb, + ) + + check = MemoryAvailability(execute_module=fake_execute_module) + result = check.run(tmp=None, task_vars=task_vars) + + assert result.get('failed', False) + for word in 'below recommended'.split() + extra_words: + assert word in result['msg'] + + +def fake_execute_module(*args): + raise AssertionError('this function should not be called') diff --git a/roles/openshift_health_checker/test/ovs_version_test.py b/roles/openshift_health_checker/test/ovs_version_test.py new file mode 100644 index 000000000..6494e1c06 --- /dev/null +++ b/roles/openshift_health_checker/test/ovs_version_test.py @@ -0,0 +1,89 @@ +import pytest + +from openshift_checks.ovs_version import OvsVersion, OpenShiftCheckException + + +def test_openshift_version_not_supported(): + def execute_module(module_name=None, module_args=None, tmp=None, task_vars=None): + return {} + + openshift_release = '111.7.0' + + task_vars = dict( + openshift=dict(common=dict(service_type='origin')), + openshift_release=openshift_release, + openshift_image_tag='v' + openshift_release, + openshift_deployment_type='origin', + ) + + check = OvsVersion(execute_module=execute_module) + with pytest.raises(OpenShiftCheckException) as excinfo: + check.run(tmp=None, task_vars=task_vars) + + assert "no recommended version of Open vSwitch" in str(excinfo.value) + + +def test_invalid_openshift_release_format(): + def execute_module(module_name=None, module_args=None, tmp=None, task_vars=None): + return {} + + task_vars = dict( + openshift=dict(common=dict(service_type='origin')), + openshift_image_tag='v0', + openshift_deployment_type='origin', + ) + + check = OvsVersion(execute_module=execute_module) + with pytest.raises(OpenShiftCheckException) as excinfo: + check.run(tmp=None, task_vars=task_vars) + assert "invalid version" in str(excinfo.value) + + +@pytest.mark.parametrize('openshift_release,expected_ovs_version', [ + ("3.5", "2.6"), + ("3.6", "2.6"), + ("3.4", "2.4"), + ("3.3", "2.4"), + ("1.0", "2.4"), +]) +def test_ovs_package_version(openshift_release, expected_ovs_version): + task_vars = dict( + openshift=dict(common=dict(service_type='origin')), + openshift_release=openshift_release, + openshift_image_tag='v' + openshift_release, + ) + return_value = object() + + def execute_module(module_name=None, module_args=None, tmp=None, task_vars=None): + assert module_name == 'rpm_version' + assert "package_list" in module_args + + for pkg in module_args["package_list"]: + if pkg["name"] == "openvswitch": + assert pkg["version"] == expected_ovs_version + + return return_value + + check = OvsVersion(execute_module=execute_module) + result = check.run(tmp=None, task_vars=task_vars) + assert result is return_value + + +@pytest.mark.parametrize('group_names,is_containerized,is_active', [ + (['masters'], False, True), + # ensure check is skipped on containerized installs + (['masters'], True, False), + (['nodes'], False, True), + (['masters', 'nodes'], False, True), + (['masters', 'etcd'], False, True), + ([], False, False), + (['etcd'], False, False), + (['lb'], False, False), + (['nfs'], False, False), +]) +def test_ovs_version_skip_when_not_master_nor_node(group_names, is_containerized, is_active): + task_vars = dict( + group_names=group_names, + openshift=dict(common=dict(is_containerized=is_containerized)), + ) + assert OvsVersion.is_active(task_vars=task_vars) == is_active diff --git a/roles/openshift_health_checker/test/package_availability_test.py b/roles/openshift_health_checker/test/package_availability_test.py index 25385339a..f7e916a46 100644 --- a/roles/openshift_health_checker/test/package_availability_test.py +++ b/roles/openshift_health_checker/test/package_availability_test.py @@ -3,6 +3,20 @@ import pytest from openshift_checks.package_availability import PackageAvailability +@pytest.mark.parametrize('pkg_mgr,is_containerized,is_active', [ + ('yum', False, True), + ('yum', True, False), + ('dnf', True, False), + ('dnf', False, False), +]) +def test_is_active(pkg_mgr, is_containerized, is_active): + task_vars = dict( + ansible_pkg_mgr=pkg_mgr, + openshift=dict(common=dict(is_containerized=is_containerized)), + ) + assert PackageAvailability.is_active(task_vars=task_vars) == is_active + + @pytest.mark.parametrize('task_vars,must_have_packages,must_not_have_packages', [ ( dict(openshift=dict(common=dict(service_type='openshift'))), diff --git a/roles/openshift_health_checker/test/package_version_test.py b/roles/openshift_health_checker/test/package_version_test.py index cc1d263bc..91eace512 100644 --- a/roles/openshift_health_checker/test/package_version_test.py +++ b/roles/openshift_health_checker/test/package_version_test.py @@ -1,21 +1,154 @@ -from openshift_checks.package_version import PackageVersion +import pytest +from openshift_checks.package_version import PackageVersion, OpenShiftCheckException -def test_package_version(): + +@pytest.mark.parametrize('openshift_release, extra_words', [ + ('111.7.0', ["no recommended version of Open vSwitch"]), + ('0.0.0', ["no recommended version of Docker"]), +]) +def test_openshift_version_not_supported(openshift_release, extra_words): + def execute_module(module_name=None, module_args=None, tmp=None, task_vars=None): + return {} + + task_vars = dict( + openshift=dict(common=dict(service_type='origin')), + openshift_release=openshift_release, + openshift_image_tag='v' + openshift_release, + openshift_deployment_type='origin', + ) + + check = PackageVersion(execute_module=execute_module) + with pytest.raises(OpenShiftCheckException) as excinfo: + check.run(tmp=None, task_vars=task_vars) + + for word in extra_words: + assert word in str(excinfo.value) + + +def test_invalid_openshift_release_format(): + def execute_module(module_name=None, module_args=None, tmp=None, task_vars=None): + return {} + + task_vars = dict( + openshift=dict(common=dict(service_type='origin')), + openshift_image_tag='v0', + openshift_deployment_type='origin', + ) + + check = PackageVersion(execute_module=execute_module) + with pytest.raises(OpenShiftCheckException) as excinfo: + check.run(tmp=None, task_vars=task_vars) + assert "invalid version" in str(excinfo.value) + + +@pytest.mark.parametrize('openshift_release', [ + "3.5", + "3.6", + "3.4", + "3.3", +]) +def test_package_version(openshift_release): + task_vars = dict( + openshift=dict(common=dict(service_type='origin')), + openshift_release=openshift_release, + openshift_image_tag='v' + openshift_release, + openshift_deployment_type='origin', + ) + return_value = object() + + def execute_module(module_name=None, module_args=None, tmp=None, task_vars=None): + assert module_name == 'aos_version' + assert "package_list" in module_args + + for pkg in module_args["package_list"]: + if "-master" in pkg["name"] or "-node" in pkg["name"]: + assert pkg["version"] == task_vars["openshift_release"] + + return return_value + + check = PackageVersion(execute_module=execute_module) + result = check.run(tmp=None, task_vars=task_vars) + assert result is return_value + + +@pytest.mark.parametrize('deployment_type,openshift_release,expected_ovs_version', [ + ("openshift-enterprise", "3.5", "2.6"), + ("origin", "3.6", "2.6"), + ("openshift-enterprise", "3.4", "2.4"), + ("origin", "3.3", "2.4"), +]) +def test_ovs_package_version(deployment_type, openshift_release, expected_ovs_version): task_vars = dict( openshift=dict(common=dict(service_type='origin')), - openshift_release='v3.5', + openshift_release=openshift_release, + openshift_image_tag='v' + openshift_release, + openshift_deployment_type=deployment_type, ) return_value = object() def execute_module(module_name=None, module_args=None, tmp=None, task_vars=None): assert module_name == 'aos_version' - assert 'prefix' in module_args - assert 'version' in module_args - assert module_args['prefix'] == task_vars['openshift']['common']['service_type'] - assert module_args['version'] == task_vars['openshift_release'] + assert "package_list" in module_args + + for pkg in module_args["package_list"]: + if pkg["name"] == "openvswitch": + assert pkg["version"] == expected_ovs_version + return return_value check = PackageVersion(execute_module=execute_module) result = check.run(tmp=None, task_vars=task_vars) assert result is return_value + + +@pytest.mark.parametrize('deployment_type,openshift_release,expected_docker_version', [ + ("origin", "3.5", "1.12"), + ("openshift-enterprise", "3.4", "1.12"), + ("origin", "3.3", "1.10"), + ("openshift-enterprise", "3.2", "1.10"), + ("origin", "3.1", "1.8"), + ("openshift-enterprise", "3.1", "1.8"), +]) +def test_docker_package_version(deployment_type, openshift_release, expected_docker_version): + task_vars = dict( + openshift=dict(common=dict(service_type='origin')), + openshift_release=openshift_release, + openshift_image_tag='v' + openshift_release, + openshift_deployment_type=deployment_type, + ) + return_value = object() + + def execute_module(module_name=None, module_args=None, tmp=None, task_vars=None): + assert module_name == 'aos_version' + assert "package_list" in module_args + + for pkg in module_args["package_list"]: + if pkg["name"] == "docker": + assert pkg["version"] == expected_docker_version + + return return_value + + check = PackageVersion(execute_module=execute_module) + result = check.run(tmp=None, task_vars=task_vars) + assert result is return_value + + +@pytest.mark.parametrize('group_names,is_containerized,is_active', [ + (['masters'], False, True), + # ensure check is skipped on containerized installs + (['masters'], True, False), + (['nodes'], False, True), + (['masters', 'nodes'], False, True), + (['masters', 'etcd'], False, True), + ([], False, False), + (['etcd'], False, False), + (['lb'], False, False), + (['nfs'], False, False), +]) +def test_package_version_skip_when_not_master_nor_node(group_names, is_containerized, is_active): + task_vars = dict( + group_names=group_names, + openshift=dict(common=dict(is_containerized=is_containerized)), + ) + assert PackageVersion.is_active(task_vars=task_vars) == is_active diff --git a/roles/openshift_health_checker/test/rpm_version_test.py b/roles/openshift_health_checker/test/rpm_version_test.py new file mode 100644 index 000000000..2f09ef965 --- /dev/null +++ b/roles/openshift_health_checker/test/rpm_version_test.py @@ -0,0 +1,82 @@ +import pytest +import rpm_version + +expected_pkgs = { + "spam": { + "name": "spam", + "version": "3.2.1", + }, + "eggs": { + "name": "eggs", + "version": "3.2.1", + }, +} + + +@pytest.mark.parametrize('pkgs, expect_not_found', [ + ( + {}, + ["spam", "eggs"], # none found + ), + ( + {"spam": ["3.2.1", "4.5.1"]}, + ["eggs"], # completely missing + ), + ( + { + "spam": ["3.2.1", "4.5.1"], + "eggs": ["3.2.1"], + }, + [], # all found + ), +]) +def test_check_pkg_found(pkgs, expect_not_found): + if expect_not_found: + with pytest.raises(rpm_version.RpmVersionException) as e: + rpm_version._check_pkg_versions(pkgs, expected_pkgs) + + assert "not found to be installed" in str(e.value) + assert set(expect_not_found) == set(e.value.problem_pkgs) + else: + rpm_version._check_pkg_versions(pkgs, expected_pkgs) + + +@pytest.mark.parametrize('pkgs, expect_not_found', [ + ( + { + 'spam': ['3.2.1'], + 'eggs': ['3.3.2'], + }, + { + "eggs": { + "required_version": "3.2", + "found_versions": ["3.3"], + } + }, # not the right version + ), + ( + { + 'spam': ['3.1.2', "3.3.2"], + 'eggs': ['3.3.2', "1.2.3"], + }, + { + "eggs": { + "required_version": "3.2", + "found_versions": ["3.3", "1.2"], + }, + "spam": { + "required_version": "3.2", + "found_versions": ["3.1", "3.3"], + } + }, # not the right version + ), +]) +def test_check_pkg_version_found(pkgs, expect_not_found): + if expect_not_found: + with pytest.raises(rpm_version.RpmVersionException) as e: + rpm_version._check_pkg_versions(pkgs, expected_pkgs) + + assert "found to be installed with an incorrect version" in str(e.value) + assert expect_not_found == e.value.problem_pkgs + else: + rpm_version._check_pkg_versions(pkgs, expected_pkgs) |