diff options
Diffstat (limited to 'roles/openshift_health_checker/openshift_checks/__init__.py')
-rw-r--r-- | roles/openshift_health_checker/openshift_checks/__init__.py | 134 |
1 files changed, 122 insertions, 12 deletions
diff --git a/roles/openshift_health_checker/openshift_checks/__init__.py b/roles/openshift_health_checker/openshift_checks/__init__.py index 02ee1d0f9..28cb53cc5 100644 --- a/roles/openshift_health_checker/openshift_checks/__init__.py +++ b/roles/openshift_health_checker/openshift_checks/__init__.py @@ -2,8 +2,11 @@ Health checks for OpenShift clusters. """ +import json import operator import os +import time +import collections from abc import ABCMeta, abstractmethod, abstractproperty from importlib import import_module @@ -27,7 +30,7 @@ class OpenShiftCheckException(Exception): class OpenShiftCheckExceptionList(OpenShiftCheckException): - """A container for multiple logging errors that may be detected in one check.""" + """A container for multiple errors that may be detected in one check.""" def __init__(self, errors): self.errors = errors super(OpenShiftCheckExceptionList, self).__init__( @@ -40,26 +43,53 @@ class OpenShiftCheckExceptionList(OpenShiftCheckException): return self.errors[index] +FileToSave = collections.namedtuple("FileToSave", "filename contents remote_filename") + + +# pylint: disable=too-many-instance-attributes; all represent significantly different state. +# Arguably they could be separated into two hashes, one for storing parameters, and one for +# storing result state; but that smells more like clutter than clarity. @six.add_metaclass(ABCMeta) class OpenShiftCheck(object): - """ - A base class for defining checks for an OpenShift cluster environment. + """A base class for defining checks for an OpenShift cluster environment. - Expect optional params: method execute_module, dict task_vars, and string tmp. + Optional init params: method execute_module, dict task_vars, and string tmp execute_module is expected to have a signature compatible with _execute_module from ansible plugins/action/__init__.py, e.g.: def execute_module(module_name=None, module_args=None, tmp=None, task_vars=None, *args): This is stored so that it can be invoked in subclasses via check.execute_module("name", args) which provides the check's stored task_vars and tmp. + + Optional init param: want_full_results + If the check can gather logs, tarballs, etc., do so when True; but no need to spend + the time if they're not wanted (won't be written to output directory). """ - def __init__(self, execute_module=None, task_vars=None, tmp=None): + def __init__(self, execute_module=None, task_vars=None, tmp=None, want_full_results=False): + # store a method for executing ansible modules from the check self._execute_module = execute_module + # the task variables and tmpdir passed into the health checker task self.task_vars = task_vars or {} self.tmp = tmp + # a boolean for disabling the gathering of results (files, computations) that won't + # actually be recorded/used + self.want_full_results = want_full_results + + # mainly for testing purposes; see execute_module_with_retries + self._module_retries = 3 + self._module_retry_interval = 5 # seconds + # state to be recorded for inspection after the check runs: + # # set to True when the check changes the host, for accurate total "changed" count self.changed = False + # list of OpenShiftCheckException for check to report (alternative to returning a failed result) + self.failures = [] + # list of FileToSave - files the check specifies to be written locally if so configured + self.files_to_save = [] + # log messages for the check - tuples of (description, msg) where msg is serializable. + # These are intended to be a sequential record of what the check observed and determined. + self.logs = [] @abstractproperty def name(self): @@ -82,7 +112,13 @@ class OpenShiftCheck(object): @abstractmethod def run(self): - """Executes a check, normally implemented as a module.""" + """Executes a check against a host and returns a result hash similar to Ansible modules. + + Actually the direction ahead is to record state in the attributes and + not bother building a result hash. Instead, return an empty hash and let + the action plugin fill it in. Or raise an OpenShiftCheckException. + Returning a hash may become deprecated if it does not prove necessary. + """ return {} @classmethod @@ -94,7 +130,43 @@ class OpenShiftCheck(object): for subclass in subclass.subclasses(): yield subclass - def execute_module(self, module_name=None, module_args=None): + def register_failure(self, error): + """Record in the check that a failure occurred. + + Recorded failures are merged into the result hash for now. They are also saved to output directory + (if provided) <check>.failures.json and registered as a log entry for context <check>.log.json. + """ + # It should be an exception; make it one if not + if not isinstance(error, OpenShiftCheckException): + error = OpenShiftCheckException(str(error)) + self.failures.append(error) + # duplicate it in the logs so it can be seen in the context of any + # information that led to the failure + self.register_log("failure: " + error.name, str(error)) + + def register_log(self, context, msg): + """Record an entry for the check log. + + Notes are intended to serve as context of the whole sequence of what the check observed. + They are be saved as an ordered list in a local check log file. + They are not to included in the result or in the ansible log; it's just for the record. + """ + self.logs.append([context, msg]) + + def register_file(self, filename, contents=None, remote_filename=""): + """Record a file that a check makes available to be saved individually to output directory. + + Either file contents should be passed in, or a file to be copied from the remote host + should be specified. Contents that are not a string are to be serialized as JSON. + + NOTE: When copying a file from remote host, it is slurped into memory as base64, meaning + you should avoid using this on huge files (more than say 10M). + """ + if contents is None and not remote_filename: + raise OpenShiftCheckException("File data/source not specified; this is a bug in the check.") + self.files_to_save.append(FileToSave(filename, contents, remote_filename)) + + def execute_module(self, module_name=None, module_args=None, save_as_name=None, register=True): """Invoke an Ansible module from a check. Invoke stored _execute_module, normally copied from the action @@ -106,6 +178,12 @@ class OpenShiftCheck(object): Ansible version). So e.g. check.execute_module("foo", dict(arg1=...)) + + save_as_name specifies a file name for saving the result to an output directory, + if needed, and is intended to uniquely identify the result of invoking execute_module. + If not provided, the module name will be used. + If register is set False, then the result won't be registered in logs or files to save. + Return: result hash from module execution. """ if self._execute_module is None: @@ -113,7 +191,33 @@ class OpenShiftCheck(object): self.__class__.__name__ + " invoked execute_module without providing the method at initialization." ) - return self._execute_module(module_name, module_args, self.tmp, self.task_vars) + result = self._execute_module(module_name, module_args, self.tmp, self.task_vars) + if result.get("changed"): + self.changed = True + for output in ["result", "stdout"]: + # output is often JSON; attempt to decode + try: + result[output + "_json"] = json.loads(result[output]) + except (KeyError, ValueError): + pass + + if register: + self.register_log("execute_module: " + module_name, result) + self.register_file(save_as_name or module_name + ".json", result) + return result + + def execute_module_with_retries(self, module_name, module_args): + """Run execute_module and retry on failure.""" + result = {} + tries = 0 + while True: + res = self.execute_module(module_name, module_args) + if tries > self._module_retries or not res.get("failed"): + result.update(res) + return result + result["last_failed"] = res + tries += 1 + time.sleep(self._module_retry_interval) def get_var(self, *keys, **kwargs): """Get deeply nested values from task_vars. @@ -171,8 +275,12 @@ class OpenShiftCheck(object): 'There is a bug in this check. While trying to convert variable \n' ' "{var}={value}"\n' 'the given converter cannot be used or failed unexpectedly:\n' - '{error}'.format(var=".".join(keys), value=value, error=error) - ) + '{type}: {error}'.format( + var=".".join(keys), + value=value, + type=error.__class__.__name__, + error=error + )) @staticmethod def get_major_minor_version(openshift_image_tag): @@ -214,7 +322,9 @@ class OpenShiftCheck(object): mount_point = os.path.dirname(mount_point) try: - return mount_for_path[mount_point] + mount = mount_for_path[mount_point] + self.register_log("mount point for " + path, mount) + return mount except KeyError: known_mounts = ', '.join('"{}"'.format(mount) for mount in sorted(mount_for_path)) raise OpenShiftCheckException( @@ -242,7 +352,7 @@ def load_checks(path=None, subpkg=""): modules = modules + load_checks(os.path.join(path, name), subpkg + "." + name) continue - if name.endswith(".py") and not name.startswith(".") and name not in LOADER_EXCLUDES: + if name.endswith(".py") and name not in LOADER_EXCLUDES: modules.append(import_module(__package__ + subpkg + "." + name[:-3])) return modules |