diff options
Diffstat (limited to 'roles/openshift_health_checker/callback_plugins')
-rw-r--r-- | roles/openshift_health_checker/callback_plugins/zz_failure_summary.py | 267 |
1 files changed, 195 insertions, 72 deletions
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 8caefab15..349655966 100644 --- a/roles/openshift_health_checker/callback_plugins/zz_failure_summary.py +++ b/roles/openshift_health_checker/callback_plugins/zz_failure_summary.py @@ -1,21 +1,22 @@ -# vim: expandtab:tabstop=4:shiftwidth=4 -''' -Ansible callback plugin. -''' +"""Ansible callback plugin to print a nicely formatted summary of failures. -from pprint import pformat +The file / module name is prefixed with `zz_` to make this plugin be loaded last +by Ansible, thus making its output the last thing that users see. +""" + +from collections import defaultdict +import traceback from ansible.plugins.callback import CallbackBase from ansible import constants as C from ansible.utils.color import stringc +FAILED_NO_MSG = u'Failed without returning a message.' + + class CallbackModule(CallbackBase): - ''' - This callback plugin stores task results and summarizes failures. - The file name is prefixed with `zz_` to make this plugin be loaded last by - Ansible, thus making its output the last thing that users see. - ''' + """This callback plugin stores task results and summarizes failures.""" CALLBACK_VERSION = 2.0 CALLBACK_TYPE = 'aggregate' @@ -25,76 +26,198 @@ class CallbackModule(CallbackBase): def __init__(self): super(CallbackModule, self).__init__() self.__failures = [] + self.__playbook_file = '' + + def v2_playbook_on_start(self, playbook): + super(CallbackModule, self).v2_playbook_on_start(playbook) + # pylint: disable=protected-access; Ansible gives us no public API to + # get the file name of the current playbook from a callback plugin. + 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)) + if not ignore_errors: + self.__failures.append(result) 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() - - def _print_failure_summary(self): - '''Print a summary of failed tasks (including ignored failures).''' - 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))) - 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): - lines = _format_failure(failure) - self._display.display(u'\n{}{}'.format(initial_indent_format.format(i), lines[0])) - for line in lines[1:]: - line = line.replace(u'\n', u'\n' + subsequent_extra_indent) - indented = u'{}{}'.format(subsequent_indent, line) - 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 -def _format_failure(failure): - '''Return a list of pretty-formatted lines describing a failure, including - relevant information about it. Line separators are not included.''' - result = failure['result'] - host = result._host.get_name() - play = _get_play(result._task) - if play: - play = play.get_name() - task = result._task.get_name() - msg = result._result.get('msg', u'???') - rows = ( - (u'Host', host), + # pylint: disable=broad-except; capturing exceptions broadly is + # intentional, to isolate arbitrary failures in this callback plugin. + try: + if self.__failures: + self._display.display(failure_summary(self.__failures, self.__playbook_file)) + except Exception: + msg = stringc( + u'An error happened while generating a summary of failures:\n' + u'{}'.format(traceback.format_exc()), C.COLOR_WARN) + self._display.v(msg) + + +def failure_summary(failures, playbook): + """Return a summary of failed tasks, including details on health checks.""" + if not failures: + return u'' + + # NOTE: because we don't have access to task_vars from callback plugins, we + # store the playbook context in the task result when the + # openshift_health_check action plugin is used, and we use this context to + # customize the error message. + # pylint: disable=protected-access; Ansible gives us no sufficient public + # API on TaskResult objects. + context = next(( + context for context in + (failure._result.get('playbook_context') for failure in failures) + if context + ), None) + + failures = [failure_to_dict(failure) for failure in failures] + failures = deduplicate_failures(failures) + + summary = [u'', u'', u'Failure summary:', u''] + + 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(failures, 1): + entries = format_failure(failure) + summary.append(u'\n{}{}'.format(initial_indent_format.format(i), entries[0])) + for entry in entries[1:]: + entry = entry.replace(u'\n', u'\n' + subsequent_extra_indent) + indented = u'{}{}'.format(subsequent_indent, entry) + summary.append(indented) + + failed_checks = set() + for failure in failures: + failed_checks.update(name for name, message in failure['checks']) + if failed_checks: + summary.append(check_failure_footer(failed_checks, context, playbook)) + + return u'\n'.join(summary) + + +def failure_to_dict(failed_task_result): + """Extract information out of a failed TaskResult into a dict. + + The intent is to transform a TaskResult object into something easier to + manipulate. TaskResult is ansible.executor.task_result.TaskResult. + """ + # pylint: disable=protected-access; Ansible gives us no sufficient public + # API on TaskResult objects. + _result = failed_task_result._result + return { + 'host': failed_task_result._host.get_name(), + 'play': play_name(failed_task_result._task), + 'task': failed_task_result.task_name, + 'msg': _result.get('msg', FAILED_NO_MSG), + 'checks': tuple( + (name, result.get('msg', FAILED_NO_MSG)) + for name, result in sorted(_result.get('checks', {}).items()) + if result.get('failed') + ), + } + + +def play_name(obj): + """Given a task or block, return the name of its parent play. + + This is loosely inspired by ansible.playbook.base.Base.dump_me. + """ + # pylint: disable=protected-access; Ansible gives us no sufficient public + # API to implement this. + if not obj: + return '' + if hasattr(obj, '_play'): + return obj._play.get_name() + return play_name(getattr(obj, '_parent')) + + +def deduplicate_failures(failures): + """Group together similar failures from different hosts. + + Returns a new list of failures such that identical failures from different + hosts are grouped together in a single entry. The relative order of failures + is preserved. + """ + groups = defaultdict(list) + for failure in failures: + group_key = tuple(sorted((key, value) for key, value in failure.items() if key != 'host')) + groups[group_key].append(failure) + result = [] + for failure in failures: + group_key = tuple(sorted((key, value) for key, value in failure.items() if key != 'host')) + if group_key not in groups: + continue + failure['host'] = tuple(sorted(g_failure['host'] for g_failure in groups.pop(group_key))) + result.append(failure) + return result + + +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 + by a newline separator when output to the user.""" + host = u', '.join(failure['host']) + play = failure['play'] + task = failure['task'] + msg = failure['msg'] + checks = failure['checks'] + fields = ( + (u'Hosts', host), (u'Play', play), (u'Task', task), (u'Message', stringc(msg, C.COLOR_ERROR)), ) - if 'checks' in result._result: - rows += ((u'Details', stringc(pformat(result._result['checks']), C.COLOR_ERROR)),) + if checks: + fields += ((u'Details', format_failed_checks(checks)),) row_format = '{:10}{}' - return [row_format.format(header + u':', body) for header, body in rows] - - -# 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 -def _get_play(obj): - '''Given a task or block, recursively tries to find its parent play.''' - if hasattr(obj, '_play'): - return obj._play - if getattr(obj, '_parent'): - return _get_play(obj._parent) + return [row_format.format(header + u':', body) for header, body in fields] + + +def format_failed_checks(checks): + """Return pretty-formatted text describing checks that failed.""" + messages = [] + for name, message in checks: + messages.append(u'check "{}":\n{}'.format(name, message)) + return stringc(u'\n\n'.join(messages), C.COLOR_ERROR) + + +def check_failure_footer(failed_checks, context, playbook): + """Return a textual explanation about checks depending on context. + + The purpose of specifying context is to vary the output depending on what + the user was expecting to happen (based on which playbook they ran). The + only use currently is to vary the message depending on whether the user was + deliberately running checks or was trying to install/upgrade and checks are + just included. Other use cases may arise. + """ + checks = ','.join(sorted(failed_checks)) + summary = [u''] + if context in ['pre-install', 'health', 'adhoc']: + # User was expecting to run checks, less explanation needed. + summary.extend([ + u'You may configure or disable checks by setting Ansible ' + u'variables. To disable those above, set:', + u' openshift_disable_check={checks}'.format(checks=checks), + u'Consult check documentation for configurable variables.', + ]) + else: + # User may not be familiar with the checks, explain what checks are in + # the first place. + summary.extend([ + u'The execution of "{playbook}" includes checks designed to fail ' + u'early if the requirements of the playbook are not met. One or ' + u'more of these checks failed. To disregard these results,' + u'explicitly disable checks by setting an Ansible variable:'.format(playbook=playbook), + u' openshift_disable_check={checks}'.format(checks=checks), + u'Failing check names are shown in the failure details above. ' + u'Some checks may be configurable by variables if your requirements ' + u'are different from the defaults; consult check documentation.', + ]) + summary.append( + u'Variables can be set in the inventory or passed on the command line ' + u'using the -e flag to ansible-playbook.' + ) + return u'\n'.join(summary) |