Python utilities.valid_string函数代码示例

  def add_relation(self, source, target, kind, label=None, metadata=None):
    """Adds a relation to the context graph."""
    assert utilities.valid_string(source) and utilities.valid_string(target)
    assert utilities.valid_string(kind)
    assert utilities.valid_optional_string(label)
    assert (metadata is None) or isinstance(metadata, types.DictType)

    with self._lock:
      # The timestamp of the relation should be inherited from the previous
      # context graph.
      key = (source, target, kind)
      timestamp = self._previous_relations_to_timestamps.get(key)
      if not utilities.valid_string(timestamp):
        timestamp = utilities.now()

      # Add the relation to the context graph data structure.
      relation = {
          'source': source,
          'target': target,
          'type': kind,
          'timestamp': timestamp
      }
      self._current_relations_to_timestamps[key] = timestamp

      # Add annotations as needed.
      relation['annotations'] = {}
      if metadata is not None:
        relation['annotations']['metadata'] = copy.deep_copy(metadata)

      relation['annotations']['label'] = label if label is not None else kind
      if self._version is not None:
        relation['annotations']['createdBy'] = self._version

      self._context_relations.append(relation)

  def test_cluster(self):
    """Test the '/cluster' endpoint."""
    start_time = utilities.now()
    # Execrcise the collector. Read data from golden files and compute
    # a context graph.
    ret_value = self.app.get('/cluster')
    end_time = utilities.now()
    result = json.loads(ret_value.data)
    self.verify_resources(result, start_time, end_time)

    self.assertEqual(23, self.count_relations(result, 'contains'))
    self.assertEqual(3, self.count_relations(result, 'createdFrom'))
    self.assertEqual(7, self.count_relations(result, 'loadBalances'))
    self.assertEqual(6, self.count_relations(result, 'monitors'))
    self.assertEqual(10, self.count_relations(result, 'runs'))

    # Verify that all relations contain a timestamp in the range
    # [start_time, end_time].
    self.assertTrue(isinstance(result.get('relations'), types.ListType))
    for r in result['relations']:
      self.assertTrue(isinstance(r, types.DictType))
      timestamp = r.get('timestamp')
      self.assertTrue(utilities.valid_string(timestamp))
      self.assertTrue(start_time <= timestamp <= end_time)

    # The overall timestamp must be in the expected range.
    self.assertTrue(utilities.valid_string(result.get('timestamp')))
    self.assertTrue(start_time <= result['timestamp'] <= end_time)

    json_output = json.dumps(result, sort_keys=True)
    self.assertEqual(2, json_output.count('"alternateLabel": '))
    self.assertEqual(85, json_output.count('"createdBy": '))

def _do_compute_pod(gs, cluster_guid, node_guid, pod, g):
  assert isinstance(gs, global_state.GlobalState)
  assert utilities.valid_string(cluster_guid)
  assert utilities.valid_string(node_guid)
  assert utilities.is_wrapped_object(pod, 'Pod')
  assert isinstance(g, ContextGraph)

  pod_id = pod['id']
  pod_guid = 'Pod:' + pod_id
  g.add_resource(pod_guid, pod['annotations'], 'Pod', pod['timestamp'],
                 pod['properties'])

  # pod.properties.spec.nodeName may be missing if the pod is waiting
  # (not running yet).
  docker_host = utilities.get_attribute(
      pod, ['properties', 'spec', 'nodeName'])
  if utilities.valid_string(docker_host):
    # Pod is running.
    if node_guid == ('Node:' + docker_host):
      g.add_relation(node_guid, pod_guid, 'runs')  # Node runs Pod
    else:
      msg = ('Docker host (pod.properties.spec.nodeName)=%s '
             'not matching node ID=%s' % (docker_host, node_guid))
      gs.logger_error(msg)
      raise collector_error.CollectorError(msg)
  else:
    # Pod is not running.
    g.add_relation(cluster_guid, pod_guid, 'contains')  # Cluster contains Pod

  def add_resource(self, rid, annotations, rtype, timestamp, obj):
    """Adds a resource to the context graph."""
    assert utilities.valid_string(rid)
    assert utilities.valid_string(utilities.get_attribute(
        annotations, ['label']))
    assert utilities.valid_string(rtype)
    assert utilities.valid_string(timestamp)
    assert isinstance(obj, types.DictType)

    with self._lock:
      # It is possible that the same resource is referenced by more than one
      # parent. In this case the resource is added only once.
      if rid in self._id_set:
        return

      # Add the resource to the context graph data structure.
      resource = {
          'id': rid,
          'type': rtype,
          'timestamp': timestamp,
          'annotations': copy.deepcopy(annotations)
      }

      if self._version is not None:
        resource['annotations']['createdBy'] = self._version

      resource['properties'] = obj

      self._context_resources.append(resource)
      self._id_set.add(rid)

  def add_relation(self, source, target, kind, label=None, metadata=None):
    """Adds a relation to the context graph."""
    assert utilities.valid_string(source) and utilities.valid_string(target)
    assert utilities.valid_string(kind)
    assert utilities.valid_optional_string(label)
    assert (metadata is None) or isinstance(metadata, types.DictType)

    with self._lock:
      # Add the relation to the context graph data structure.
      relation = {
          'source': source,
          'target': target,
          'type': kind,
      }

      # Add annotations as needed.
      relation['annotations'] = {}
      if metadata is not None:
        relation['annotations']['metadata'] = metadata

      relation['annotations']['label'] = label if label is not None else kind
      if self._version is not None:
        relation['annotations']['createdBy'] = self._version

      self._context_relations.append(relation)

def _get_container_labels(container, parent_pod):
    """Returns key/value pairs identifying all metrics of this container.

  Args:
    container: the container object to annotate.
    parent_pod: the parent pod of 'container'.

  Returns:
  A dictionary of key/value pairs.
  If any error was detected, returns None.
  """
    if not utilities.is_wrapped_object(container, "Container"):
        return None
    if not utilities.is_wrapped_object(parent_pod, "Pod"):
        return None

    pod_id = utilities.get_attribute(parent_pod, ["properties", "metadata", "uid"])
    if not utilities.valid_string(pod_id):
        return None

    hostname = utilities.get_attribute(parent_pod, ["properties", "spec", "host"])
    if not utilities.valid_string(hostname):
        return None

    short_container_name = utilities.get_short_container_name(container, parent_pod)

    if not utilities.valid_string(short_container_name):
        return None

    return {"pod_id": pod_id, "hostname": hostname, "container_name": short_container_name}

  def best_label(self, obj):
    """Returns the best human-readable label of the given object.

    We perfer the "alternateLabel" over "label" and a string not composed
    of only hexadecimal digits over hexadecimal digits.

    This function must be called when self._lock is held.

    Args:
      obj: a dictionary containing an "annotations" attribute. The value
        of this attribute should be a dictionary, which may contain
        "alternateLabel" and "Label" attributes.

    Returns:
    The best human-readable label.
    """
    alt_label = utilities.get_attribute(obj, ['annotations', 'alternateLabel'])
    label = utilities.get_attribute(obj, ['annotations', 'label'])
    if (utilities.valid_string(alt_label) and
        re.search('[^0-9a-fA-F]', alt_label)):
      return alt_label
    elif utilities.valid_string(label) and re.search('[^0-9a-fA-F]', label):
      return label
    elif utilities.valid_string(alt_label):
      return alt_label
    elif utilities.valid_string(label):
      return label
    else:
      # should not arrive here.
      return '<unknown>'

def get_pod_host(gs, pod_id):
  """Gets the host name associated with the given pod.

  Args:
    gs: global state.
    pod_id: the pod name.

  Returns:
    If the pod was found, returns the associated host name.
    If the pod was not found, returns an empty string.

  Raises:
    CollectorError in case of failure to fetch data from Kubernetes.
    Other exceptions may be raised due to exectution errors.
  """
  gs.logger_info('calling get_pod_host(pod_id=%s)', pod_id)
  for pod in get_pods(gs):
    if not utilities.valid_string(pod.get('id')):
      # Found an invalid pod without a pod ID.
      continue

    pod_host = utilities.get_attribute(pod, ['properties', 'spec', 'host'])
    if pod['id'] == pod_id and utilities.valid_string(pod_host):
      # 'pod_host' may be missing if the pod is in "Waiting" state.
      return pod_host

  # Could not find pod.
  return ''

def annotate_container(project_id, container, parent_pod):
    """Annotate the given container with Heapster GCM metric information.

  Args:
    project_id: the project ID
    container: the container object to annotate.
    parent_pod: the parent pod of 'container'.

  Raises:
    AssertionError: if the input arguments are invalid or if
    'parent_pod' is not the parent of 'container'
  """
    assert utilities.valid_string(project_id)
    assert utilities.is_wrapped_object(container, "Container")
    assert utilities.is_wrapped_object(parent_pod, "Pod")
    parent_name = utilities.get_attribute(container, ["properties", "Config", "Hostname"])
    assert utilities.valid_string(parent_name)
    pod_name = utilities.get_attribute(parent_pod, ["properties", "metadata", "name"])
    assert utilities.valid_string(pod_name)

    # The 'parent_name' value is truncated to the first 64 characters.
    # Thus it must be the prefix of the full pod name.
    assert pod_name.startswith(parent_name)

    m = _make_gcm_metrics(project_id, _get_container_labels(container, parent_pod))
    if m is None:
        return
    if container.get("annotations") is None:
        container["annotations"] = {}
    container["annotations"]["metrics"] = m

def _do_compute_node(gs, input_queue, cluster_guid, node, g):
  assert isinstance(gs, global_state.GlobalState)
  assert isinstance(input_queue, Queue.PriorityQueue)
  assert utilities.valid_string(cluster_guid)
  assert utilities.is_wrapped_object(node, 'Node')
  assert isinstance(g, ContextGraph)

  node_id = node['id']
  node_guid = 'Node:' + node_id
  g.add_resource(node_guid, node['annotations'], 'Node', node['timestamp'],
                 node['properties'])
  g.add_relation(cluster_guid, node_guid, 'contains')  # Cluster contains Node
  # Pods in a Node
  pod_ids = set()
  docker_hosts = set()

  # Process pods sequentially because calls to _do_compute_pod() do not call
  # lower-level services or wait.
  for pod in kubernetes.get_pods(gs, node_id):
    _do_compute_pod(gs, cluster_guid, node_guid, pod, g)
    pod_ids.add(pod['id'])
    # pod.properties.spec.nodeName may be missing if the pod is waiting.
    docker_host = utilities.get_attribute(
        pod, ['properties', 'spec', 'nodeName'])
    if utilities.valid_string(docker_host):
      docker_hosts.add(docker_host)

  # 'docker_hosts' should contain a single Docker host, because all of
  # the pods run in the same Node. However, if it is not the case, we
  # cannot fix the situation, so we just log an error message and continue.
  if len(docker_hosts) != 1:
    gs.logger_error(
        'corrupt pod data in node=%s: '
        '"docker_hosts" is empty or contains more than one entry: %s',
        node_guid, str(docker_hosts))

  # Process containers concurrently.
  for docker_host in docker_hosts:
    for container in docker.get_containers_with_metrics(gs, docker_host):
      parent_pod_id = utilities.get_parent_pod_id(container)
      if utilities.valid_string(parent_pod_id) and (parent_pod_id in pod_ids):
        # This container is contained in a pod.
        parent_guid = 'Pod:' + parent_pod_id
      else:
        # This container is not contained in a pod.
        parent_guid = node_guid

      # Do not compute the containers by worker threads in test mode
      # because the order of the output will be different than the golden
      # files due to the effects of queuing the work.
      if gs.get_testing():
        _do_compute_container(gs, docker_host, parent_guid, container, g)
      else:
        input_queue.put((
            gs.get_random_priority(),
            _do_compute_container,
            {'gs': gs, 'docker_host': docker_host, 'parent_guid': parent_guid,
             'container': container, 'g': g}))

def _do_compute_other_nodes(gs, cluster_guid, nodes_list, oldest_timestamp, g):
  """Adds nodes not in the node list but running pods to the graph.

  This handles the case when there are pods running on the master node,
  in which case we add a dummy node representing the master to the graph.
  The nodes list does not include the master.

  Args:
    gs: the global state.
    cluster_guid: the cluster's ID.
    nodes_list: a list of wrapped Node objects.
    oldest_timestamp: the timestamp of the oldest Node object.
    g: the context graph under construction.
  """
  assert isinstance(gs, global_state.GlobalState)
  assert utilities.valid_string(cluster_guid)
  assert isinstance(nodes_list, list)
  assert utilities.valid_string(oldest_timestamp)
  assert isinstance(g, ContextGraph)

  # Compute the set of known Node names.
  known_node_ids = set()
  for node in nodes_list:
    assert utilities.is_wrapped_object(node, 'Node')
    known_node_ids.add(node['id'])

  # Compute the set of Nodes referenced by pods but not in the known set.
  # The set of unknown node names may be empty.
  missing_node_ids = set()
  for pod in kubernetes.get_pods(gs):
    assert utilities.is_wrapped_object(pod, 'Pod')
    # pod.properties.spec.nodeName may be missing if the pod is waiting.
    parent_node_id = utilities.get_attribute(
        pod, ['properties', 'spec', 'nodeName'])
    if not utilities.valid_string(parent_node_id):
      continue

    if parent_node_id in known_node_ids:
      continue

    # Found a pod that does not belong to any of the known nodes.
    missing_node_ids.add(parent_node_id)

  # Process the missing nodes.
  for node_id in missing_node_ids:
    # Create a dummy node object just as a placeholder for metric
    # annotations.
    node = utilities.wrap_object({}, 'Node', node_id, time.time())

    metrics.annotate_node(node)
    node_guid = 'Node:' + node_id
    g.add_resource(node_guid, node['annotations'], 'Node', oldest_timestamp, {})
    g.add_relation(cluster_guid, node_guid, 'contains')  # Cluster contains Node

def _do_compute_node(gs, input_queue, cluster_guid, node, g):
  assert isinstance(gs, global_state.GlobalState)
  assert isinstance(input_queue, Queue.PriorityQueue)
  assert utilities.valid_string(cluster_guid)
  assert utilities.is_wrapped_object(node, 'Node')
  assert isinstance(g, ContextGraph)

  node_id = node['id']
  node_guid = 'Node:' + node_id
  g.add_resource(node_guid, node['annotations'], 'Node', node['timestamp'],
                 node['properties'])
  g.add_relation(cluster_guid, node_guid, 'contains')  # Cluster contains Node
  # Pods in a Node
  # Do not compute the pods by worker threads in test mode because the order
  # of the output will be different than the golden files due to the effects
  # of queuing the work.
  for pod in kubernetes.get_pods(gs, node_id):
    if gs.get_testing():
      _do_compute_pod(gs, input_queue, node_guid, pod, g)
    else:
      input_queue.put((
          gs.get_random_priority(),
          _do_compute_pod,
          {'gs': gs, 'input_queue': input_queue, 'node_guid': node_guid,
           'pod': pod, 'g': g}))

def _do_compute_service(gs, cluster_guid, service, g):
  assert isinstance(gs, global_state.GlobalState)
  assert utilities.valid_string(cluster_guid)
  assert utilities.is_wrapped_object(service, 'Service')
  assert isinstance(g, ContextGraph)

  service_id = service['id']
  service_guid = 'Service:' + service_id
  g.add_resource(service_guid, service['annotations'], 'Service',
                 service['timestamp'], service['properties'])

  # Cluster contains Service.
  g.add_relation(cluster_guid, service_guid, 'contains')

  # Pods load balanced by this service (use the service['spec', 'selector']
  # key/value pairs to find matching Pods)
  selector = utilities.get_attribute(
      service, ['properties', 'spec', 'selector'])
  if selector:
    if not isinstance(selector, types.DictType):
      msg = 'Service id=%s has an invalid "selector" value' % service_id
      gs.logger_error(msg)
      raise collector_error.CollectorError(msg)

    for pod in kubernetes.get_selected_pods(gs, selector):
      pod_guid = 'Pod:' + pod['id']
      # Service loadBalances Pod
      g.add_relation(service_guid, pod_guid, 'loadBalances')
  else:
    gs.logger_error('Service id=%s has no "selector" attribute', service_id)

def get_pods(gs, node_id=None):
  """Gets the list of all pods in the given node or in the cluster.

  When 'node_id' is None, it returns the list of pods in the cluster.
  When 'node_id' is a non-empty string, it returns the list of pods in that
  node.

  Args:
    gs: global state.
    node_id: the parent node of the pods or None.

  Returns:
    list of wrapped pod objects.
    Each element in the list is the result of
    utilities.wrap_object(pod, 'Pod', ...)

  Raises:
    CollectorError: in case of failure to fetch data from Kubernetes.
    Other exceptions may be raised due to exectution errors.
  """
  pods_label = '' if node_id is None else node_id
  pods, timestamp_secs = gs.get_pods_cache().lookup(pods_label)
  if timestamp_secs is not None:
    gs.logger_info('get_pods(pods_label=%s) cache hit returns %d pods',
                   pods_label, len(pods))
    return pods

  pods = []
  url = '{kubernetes}/pods'.format(kubernetes=KUBERNETES_API)
  try:
    result = fetch_data(gs, url)
  except:
    msg = 'fetching %s failed with exception %s' % (url, sys.exc_info()[0])
    gs.logger_exception(msg)
    raise collector_error.CollectorError(msg)

  now = time.time()
  if not (isinstance(result, types.DictType) and 'items' in result):
    msg = 'invalid result when fetching %s' % url
    gs.logger_exception(msg)
    raise collector_error.CollectorError(msg)

  for pod in result['items']:
    name = utilities.get_attribute(pod, ['metadata', 'name'])
    if not utilities.valid_string(name):
      # an invalid pod without a valid pod ID value.
      continue
    wrapped_pod = utilities.wrap_object(pod, 'Pod', name, now)
    if node_id:
      # pod['spec']['host'] may be missing if the pod is in "Waiting"
      # status.
      if utilities.get_attribute(pod, ['spec', 'host']) == node_id:
        pods.append(wrapped_pod)
    else:
      # append pod to output if 'node_id' is not specified.
      pods.append(wrapped_pod)

  ret_value = gs.get_pods_cache().update(pods_label, pods, now)
  gs.logger_info('get_pods(node_id=%s) returns %d pods', pods_label, len(pods))
  return ret_value

def fetch(req):
  """Fetch the output of the specified request from the Docker's socket.

  Args:
    req: the request to be sent to the Docker daemon.

  Returns:
  The contents of the JSON response.

  Raises:
    IOError: the Unix domain socket returns a code other than OK (200).
  """
  assert utilities.valid_string(req)
  assert req[0] == '/'
  if app.proxy_is_testing_mode:
    fname = 'testdata/localhost' + re.sub(r'[^a-zA-Z0-9_.-]', '.', req)
    app.logger.info('reading req %s from %s', req, fname)
    f = open(fname, 'r')
    result = json.loads(f.read())
    f.close()
    return result

  r = session.get(
      '{docker_host}{url}'.format(docker_host=LOCAL_DOCKER_HOST, url=req))

  if r.status_code != requests.codes.ok:
    msg = 'Accessing %s API returns an error code %d' % (req, r.status_code)
    app.logger.error(msg)
    raise IOError(msg)

  else:
    return r.json()