def get_nodes():
"""Computes the response of the '/cluster/resources/nodes' endpoint.
Returns:
The nodes of the context graph.
"""
gs = app.context_graph_global_state
try:
nodes_list = kubernetes.get_nodes_with_metrics(gs)
except collector_error.CollectorError as e:
return flask.jsonify(utilities.make_error(str(e)))
return flask.jsonify(utilities.make_response(nodes_list, 'resources'))
def get_nodes():
"""Computes the response of the '/cluster/resources/nodes' endpoint.
Returns:
The nodes of the context graph.
"""
gs = app.context_graph_global_state
try:
nodes_list = kubernetes.get_nodes_with_metrics(gs)
except collector_error.CollectorError as e:
return flask.jsonify(utilities.make_error(str(e)))
return flask.jsonify(utilities.make_response(nodes_list, 'resources'))
def get_nodes():
"""Computes the response of the '/cluster/resources/nodes' endpoint.
Returns:
The nodes of the context graph.
"""
gs = app.context_graph_global_state
try:
nodes_list = kubernetes.get_nodes_with_metrics(gs)
except collector_error.CollectorError as e:
return flask.jsonify(make_error(str(e)))
except:
msg = 'kubernetes.get_nodes() failed with exception %s' % sys.exc_info()[0]
app.logger.exception(msg)
return flask.jsonify(make_error(msg))
return flask.jsonify(make_response(nodes_list, 'resources'))
def _do_compute_graph(gs, input_queue, output_queue, output_format):
"""Returns the context graph in the specified format.
Args:
gs: the global state.
input_queue: the input queue for the worker threads.
output_queue: output queue containing exceptions data from the worker
threads.
output_format: one of 'graph', 'dot', 'context_graph', or 'resources'.
Returns:
A successful response in the specified format.
Raises:
CollectorError: inconsistent or invalid graph data.
"""
assert isinstance(gs, global_state.GlobalState)
assert isinstance(input_queue, Queue.PriorityQueue)
assert isinstance(output_queue, Queue.Queue)
assert utilities.valid_string(output_format)
g = ContextGraph()
g.set_version(docker.get_version(gs))
g.set_metadata({'timestamp': datetime.datetime.now().isoformat()})
# Nodes
nodes_list = kubernetes.get_nodes_with_metrics(gs)
if not nodes_list:
return g.dump(gs, output_format)
# Get the cluster name from the first node.
# The cluster name is an approximation. It is not a big deal if it
# is incorrect, since the aggregator knows the cluster name.
cluster_name = utilities.node_id_to_cluster_name(nodes_list[0]['id'])
cluster_guid = 'Cluster:' + cluster_name
g.set_title(cluster_name)
g.add_resource(cluster_guid, {'label': cluster_name}, 'Cluster',
nodes_list[0]['timestamp'], {})
# Nodes
for node in nodes_list:
input_queue.put((
gs.get_random_priority(),
_do_compute_node,
{'gs': gs, 'input_queue': input_queue, 'cluster_guid': cluster_guid,
'node': node, 'g': g}))
# Services
for service in kubernetes.get_services(gs):
input_queue.put((
gs.get_random_priority(),
_do_compute_service,
{'gs': gs, 'cluster_guid': cluster_guid, 'service': service, 'g': g}))
# ReplicationControllers
rcontrollers_list = kubernetes.get_rcontrollers(gs)
for rcontroller in rcontrollers_list:
input_queue.put((
gs.get_random_priority(),
_do_compute_rcontroller,
{'gs': gs, 'cluster_guid': cluster_guid, 'rcontroller': rcontroller,
'g': g}))
# Wait until worker threads finished processing all outstanding requests.
# Once we return from the join(), all output was generated already.
input_queue.join()
# Convert any exception caught by the worker threads to an exception
# raised by the current thread.
if not output_queue.empty():
msg = output_queue.get_nowait() # should not fail.
gs.logger_error(msg)
raise collector_error.CollectorError(msg)
# Dump the resulting graph
return g.dump(gs, output_format)
def _do_compute_graph(gs, input_queue, output_queue, output_format):
"""Returns the context graph in the specified format.
Args:
gs: the global state.
input_queue: the input queue for the worker threads.
output_queue: output queue containing exceptions data from the worker
threads.
output_format: one of 'graph', 'dot', 'context_graph', or 'resources'.
Returns:
A successful response in the specified format.
Raises:
CollectorError: inconsistent or invalid graph data.
"""
assert isinstance(gs, global_state.GlobalState)
assert isinstance(input_queue, Queue.PriorityQueue)
assert isinstance(output_queue, Queue.Queue)
assert utilities.valid_string(output_format)
g = ContextGraph()
g.set_version(docker.get_version(gs))
g.set_metadata({'timestamp': utilities.now()})
g.set_relations_to_timestamps(gs.get_relations_to_timestamps())
# Nodes
nodes_list = kubernetes.get_nodes_with_metrics(gs)
if not nodes_list:
return g.dump(gs, output_format)
# Find the timestamp of the oldest node. This will be the timestamp of
# the cluster.
oldest_timestamp = utilities.now()
for node in nodes_list:
assert utilities.is_wrapped_object(node, 'Node')
# note: we cannot call min(oldest_timestamp, node['timestamp']) here
# because min(string) returnes the smallest character in the string.
if node['timestamp'] < oldest_timestamp:
oldest_timestamp = node['timestamp']
# Get the cluster name from the first node.
# The cluster name is an approximation. It is not a big deal if it
# is incorrect, since the aggregator knows the cluster name.
cluster_name = utilities.node_id_to_cluster_name(nodes_list[0]['id'])
cluster_guid = 'Cluster:' + cluster_name
g.set_title(cluster_name)
g.add_resource(cluster_guid, {'label': cluster_name}, 'Cluster',
oldest_timestamp, {})
# Nodes
for node in nodes_list:
input_queue.put((gs.get_random_priority(), _do_compute_node, {
'gs': gs,
'input_queue': input_queue,
'cluster_guid': cluster_guid,
'node': node,
'g': g
}))
# Services
for service in kubernetes.get_services(gs):
input_queue.put((gs.get_random_priority(), _do_compute_service, {
'gs': gs,
'cluster_guid': cluster_guid,
'service': service,
'g': g
}))
# ReplicationControllers
rcontrollers_list = kubernetes.get_rcontrollers(gs)
for rcontroller in rcontrollers_list:
input_queue.put((gs.get_random_priority(), _do_compute_rcontroller, {
'gs': gs,
'cluster_guid': cluster_guid,
'rcontroller': rcontroller,
'g': g
}))
# Wait until worker threads finished processing all outstanding requests.
# Once we return from the join(), all output was generated already.
input_queue.join()
# Convert any exception caught by the worker threads to an exception
# raised by the current thread.
if not output_queue.empty():
msg = output_queue.get_nowait() # should not fail.
gs.logger_error(msg)
raise collector_error.CollectorError(msg)
# Keep the relations_to_timestamps mapping for next call.
gs.set_relations_to_timestamps(g.get_relations_to_timestamps())
# Dump the resulting graph
return g.dump(gs, output_format)
def _do_compute_graph(gs, output_format):
"""Returns the context graph in the specified format.
Args:
gs: the global state.
output_format: one of 'dot', 'context_graph', or 'resources'.
Returns:
A successful response in the specified format.
Raises:
CollectorError: inconsistent or invalid graph data.
"""
assert isinstance(gs, global_state.GlobalState)
assert utilities.valid_string(output_format)
g = ContextGraph()
g.set_relations_to_timestamps(gs.get_relations_to_timestamps())
# Nodes
nodes_list = kubernetes.get_nodes_with_metrics(gs)
if not nodes_list:
return g.dump(output_format)
# Find the timestamp of the oldest node. This will be the timestamp of
# the cluster.
oldest_timestamp = utilities.now()
for node in nodes_list:
assert utilities.is_wrapped_object(node, 'Node')
# note: we cannot call min(oldest_timestamp, node['timestamp']) here
# because min(string) returnes the smallest character in the string.
if node['timestamp'] < oldest_timestamp:
oldest_timestamp = node['timestamp']
# The cluster name may be available through the Kubernetes API someday.
# TODO(rimey): Determine the cluster name.
cluster_name = '_unknown_'
cluster_guid = 'Cluster:' + cluster_name
g.set_title(cluster_name)
g.add_resource(cluster_guid, {'label': cluster_name}, 'Cluster',
oldest_timestamp, {})
# Nodes
for node in nodes_list:
_do_compute_node(cluster_guid, node, g)
# Pods
for pod in kubernetes.get_pods(gs):
_do_compute_pod(cluster_guid, pod, g)
# Services
for service in kubernetes.get_services(gs):
_do_compute_service(gs, cluster_guid, service, g)
# ReplicationControllers
for rcontroller in kubernetes.get_rcontrollers(gs):
_do_compute_rcontroller(gs, cluster_guid, rcontroller, g)
# Other nodes, not on the list, such as the Kubernetes master.
_do_compute_other_nodes(gs, cluster_guid, nodes_list, oldest_timestamp, g)
# Keep the relations_to_timestamps mapping for next call.
gs.set_relations_to_timestamps(g.get_relations_to_timestamps())
g.set_metadata({'timestamp': g.max_resources_and_relations_timestamp()})
# Dump the resulting graph
return g.dump(output_format)
def _do_compute_graph(gs, input_queue, output_queue, output_format):
"""Returns the context graph in the specified format.
Args:
gs: the global state.
input_queue: the input queue for the worker threads.
output_queue: output queue containing exceptions data from the worker
threads.
output_format: one of 'graph', 'dot', 'context_graph', or 'resources'.
Returns:
A successful response in the specified format.
Raises:
CollectorError: inconsistent or invalid graph data.
"""
assert isinstance(gs, global_state.GlobalState)
assert isinstance(input_queue, Queue.PriorityQueue)
assert isinstance(output_queue, Queue.Queue)
assert utilities.valid_string(output_format)
g = ContextGraph()
try:
version = docker.get_version(gs)
except Exception as e:
exc_type, value, _ = sys.exc_info()
msg = ('get_version() failed with exception %s: %s' %
(exc_type, value))
gs.logger_error(msg)
version = '_unknown_'
g.set_version(version)
g.set_relations_to_timestamps(gs.get_relations_to_timestamps())
# Nodes
nodes_list = kubernetes.get_nodes_with_metrics(gs)
if not nodes_list:
return g.dump(gs, output_format)
# Find the timestamp of the oldest node. This will be the timestamp of
# the cluster.
oldest_timestamp = utilities.now()
for node in nodes_list:
assert utilities.is_wrapped_object(node, 'Node')
# note: we cannot call min(oldest_timestamp, node['timestamp']) here
# because min(string) returnes the smallest character in the string.
if node['timestamp'] < oldest_timestamp:
oldest_timestamp = node['timestamp']
# Get the cluster name from the first node.
# The cluster name is an approximation. It is not a big deal if it
# is incorrect, since the aggregator knows the cluster name.
cluster_name = utilities.node_id_to_cluster_name(nodes_list[0]['id'])
cluster_guid = 'Cluster:' + cluster_name
g.set_title(cluster_name)
g.add_resource(cluster_guid, {'label': cluster_name}, 'Cluster',
oldest_timestamp, {})
# Nodes
for node in nodes_list:
input_queue.put((
gs.get_random_priority(),
_do_compute_node,
{'gs': gs, 'input_queue': input_queue, 'cluster_guid': cluster_guid,
'node': node, 'g': g}))
# Services
for service in kubernetes.get_services(gs):
input_queue.put((
gs.get_random_priority(),
_do_compute_service,
{'gs': gs, 'cluster_guid': cluster_guid, 'service': service, 'g': g}))
# ReplicationControllers
rcontrollers_list = kubernetes.get_rcontrollers(gs)
for rcontroller in rcontrollers_list:
input_queue.put((
gs.get_random_priority(),
_do_compute_rcontroller,
{'gs': gs, 'cluster_guid': cluster_guid, 'rcontroller': rcontroller,
'g': g}))
# Pods running on the master node.
input_queue.put((
gs.get_random_priority(),
_do_compute_master_pods,
{'gs': gs, 'cluster_guid': cluster_guid, 'nodes_list': nodes_list,
'oldest_timestamp': oldest_timestamp, 'g': g}))
# Wait until worker threads finished processing all outstanding requests.
# Once we return from the join(), all output was generated already.
input_queue.join()
# Convert any exception caught by the worker threads to an exception
# raised by the current thread.
if not output_queue.empty():
msg = output_queue.get_nowait() # should not fail.
gs.logger_error(msg)
raise collector_error.CollectorError(msg)
# Keep the relations_to_timestamps mapping for next call.
gs.set_relations_to_timestamps(g.get_relations_to_timestamps())
g.set_metadata({'timestamp': g.max_resources_and_relations_timestamp()})
# Dump the resulting graph
return g.dump(gs, output_format)