def start_arena(unit_id): g_logger = log_client.logger('arena-actions') g_logger.log_struct({"message": "Starting arena {}".format(unit_id)}, severity=LOG_LEVELS.INFO) unit = ds_client.get(ds_client.key('cybergym-unit', unit_id)) state_transition(entity=unit, new_state=BUILD_STATES.STARTING) unit['arena']['running'] = True unit['arena']['gm_start_time'] = str(calendar.timegm(time.gmtime())) ds_client.put(unit) # Start the central servers g_logger.log_struct( {"message": "Starting central servers for arena {}".format(unit_id)}, severity=LOG_LEVELS.INFO) query_central_arena_servers = ds_client.query(kind='cybergym-server') query_central_arena_servers.add_filter("workout", "=", unit_id) for server in list(query_central_arena_servers.fetch()): # Publish to a server management topic pubsub_topic = PUBSUB_TOPICS.MANAGE_SERVER publisher = pubsub_v1.PublisherClient() topic_path = publisher.topic_path(project, pubsub_topic) future = publisher.publish(topic_path, data=b'Server Build', server_name=server['name'], action=SERVER_ACTIONS.START) print(future.result()) # Now start all of the student workouts for this arena for workout_id in unit['workouts']: start_vm(workout_id)
def rebuild_workout(workout_id): """ Builds a full workout :param workout_id: The workout_id to build """ workout = ds_client.get(ds_client.key('cybergym-workout', workout_id)) state_transition(entity=workout, new_state=BUILD_STATES.START) build_workout(workout_id) return
def check_build_state_change(build_id, check_server_state, change_build_state): query_workout_servers = ds_client.query(kind='cybergym-server') query_workout_servers.add_filter("workout", "=", build_id) for check_server in list(query_workout_servers.fetch()): if check_server['state'] != check_server_state: return # If we've made it this far, then all of the servers have changed to the desired state. # now we can change the entire state. build = ds_client.get(ds_client.key('cybergym-workout', build_id)) if not build: build = ds_client.get(ds_client.key('cybergym-unit', build_id)) state_transition(build, change_build_state)
def start_vm(workout_id): print("Starting workout %s" % workout_id) workout = ds_client.get(ds_client.key('cybergym-workout', workout_id)) state_transition(entity=workout, new_state=BUILD_STATES.STARTING) workout['start_time'] = str(calendar.timegm(time.gmtime())) ds_client.put(workout) query_workout_servers = ds_client.query(kind='cybergym-server') query_workout_servers.add_filter("workout", "=", workout_id) for server in list(query_workout_servers.fetch()): # Publish to a server management topic pubsub_topic = PUBSUB_TOPICS.MANAGE_SERVER publisher = pubsub_v1.PublisherClient() topic_path = publisher.topic_path(project, pubsub_topic) future = publisher.publish(topic_path, data=b'Server Build', server_name=server['name'], action=SERVER_ACTIONS.START) print(future.result())
def nuke_workout(workout_id): """ :param workout_id: The ID of the workout specification in the Datastore :returns: None """ g_logger = log_client.logger(str(workout_id)) workout = ds_client.get(ds_client.key('cybergym-workout', workout_id)) g_logger.log_text("Nuke Operation: Delete") state_transition(entity=workout, new_state=BUILD_STATES.NUKING) DeletionManager(deletion_type=DeletionManager.DeletionType.SPECIFIC, build_id=workout_id, build_type=WORKOUT_TYPES.WORKOUT).run() time.sleep(60) g_logger.log_text("Nuke Operation: Rebuild") state_transition(entity=workout, new_state=BUILD_STATES.START) build_workout(workout_id) return
def stop_workout(workout_id): result = compute.instances().list( project=project, zone=zone, filter='name = {}*'.format(workout_id)).execute() workout = ds_client.get(ds_client.key('cybergym-workout', workout_id)) state_transition(entity=workout, new_state=BUILD_STATES.READY, existing_state=BUILD_STATES.RUNNING) ds_client.put(workout) if 'items' in result: for vm_instance in result['items']: response = compute.instances().stop( project=project, zone=zone, instance=vm_instance["name"]).execute() print("Workouts stopped") else: print("No workouts to stop")
def register_student_entry(build_id, server_name): build = ds_client.get(ds_client.key('cybergym-workout', build_id)) if not build: build = ds_client.get(ds_client.key('cybergym-unit', build_id)) # Add the external_IP address for the workout. This allows easy deletion of the DNS record when deleting the arena ip_address = get_server_ext_address(server_name) add_dns_record(build_id, ip_address) # Make sure the guacamole server actually comes up successfully before setting the workout state to ready print(f"DNS record set for {server_name}. Now Testing guacamole connection. This may take a few minutes.") if test_guacamole(ip_address): # Now, since this is the guacamole server, update the state of the workout to READY print(f"Setting the build {build_id} to ready") state_transition(entity=build, new_state=BUILD_STATES.READY) else: state_transition(entity=build, new_state=BUILD_STATES.GUACAMOLE_SERVER_LOAD_TIMEOUT) # Return the IP address used for the server build function to set the server datastore element return ip_address
def stop_workout(workout_id): result = compute.instances().list( project=project, zone=zone, filter='name = {}*'.format(workout_id)).execute() workout = ds_client.get(ds_client.key('cybergym-workout', workout_id)) state_transition(entity=workout, new_state=BUILD_STATES.READY, existing_state=BUILD_STATES.RUNNING) start_time = None if 'start_time' in workout: start_time = workout['start_time'] stop_time = calendar.timegm(time.gmtime()) runtime = int(stop_time) - int(start_time) if 'runtime_counter' in workout: accumulator = workout['runtime_counter'] new_runtime = int(accumulator) + runtime workout['runtime_counter'] = new_runtime else: workout['runtime_counter'] = runtime ds_client.put(workout) query_workout_servers = ds_client.query(kind='cybergym-server') query_workout_servers.add_filter("workout", "=", workout_id) for server in list(query_workout_servers.fetch()): # Publish to a server management topic pubsub_topic = PUBSUB_TOPICS.MANAGE_SERVER publisher = pubsub_v1.PublisherClient() topic_path = publisher.topic_path(project, pubsub_topic) future = publisher.publish(topic_path, data=b'Server Build', server_name=server['name'], action=SERVER_ACTIONS.STOP) print(future.result()) g_logger = log_client.logger(str(workout_id)) if 'items' in result: for vm_instance in result['items']: response = compute.instances().stop( project=project, zone=zone, instance=vm_instance["name"]).execute() g_logger.log_struct({"message": "Workout stopped"}, severity=LOG_LEVELS.INFO) else: g_logger.log_struct({"message": "No workouts to stop"}, severity=LOG_LEVELS.WARNING)
def server_stop(server_name): server = ds_client.get(ds_client.key('cybergym-server', server_name)) g_logger = log_client.logger(str(server_name)) state_transition(entity=server, new_state=SERVER_STATES.STOPPING) i = 0 stop_success = False while not stop_success and i < 5: workout_globals.refresh_api() try: response = compute.instances().stop( project=project, zone=zone, instance=server_name).execute() stop_success = True g_logger.log_text( f'Sent job to start {server_name}, and waiting for response') return True except BrokenPipeError: i += 1 return False
def server_delete(server_name): server = ds_client.get(ds_client.key('cybergym-server', server_name)) state_transition(entity=server, new_state=SERVER_STATES.DELETING) workout_globals.refresh_api() try: response = compute.instances().delete(project=project, zone=zone, instance=server_name).execute() except HttpError as exception: # If the server is already deleted or no longer exists, state_transition(entity=server, new_state=SERVER_STATES.DELETED) print(f"Finished deleting {server_name}") # If all servers in the workout have been deleted, then set the workout state to True build_id = server['workout'] check_build_state_change(build_id=build_id, check_server_state=SERVER_STATES.DELETED, change_build_state=BUILD_STATES.COMPLETED_DELETING_SERVERS) return True print(f'Sent delete request to {server_name}, and waiting for response') i = 0 success = False while not success and i < 5: try: print(f"Begin waiting for delete response from operation {response['id']}") compute.zoneOperations().wait(project=project, zone=zone, operation=response["id"]).execute() success = True except timeout: i += 1 print('Response timeout for deleting server. Trying again') pass if not success: print(f'Timeout in trying to delete server {server_name}') state_transition(entity=server, new_state=SERVER_STATES.BROKEN) return False # If this is a student entry server, delete the DNS if 'student_entry' in server and server['student_entry']: print(f'Deleting DNS record for {server_name}') ip_address = server['external_ip'] delete_dns(server['workout'], ip_address) state_transition(entity=server, new_state=SERVER_STATES.DELETED) print(f"Finished deleting {server_name}") # If all servers in the workout have been deleted, then set the workout state to True build_id = server['workout'] check_build_state_change(build_id=build_id, check_server_state=SERVER_STATES.DELETED, change_build_state=BUILD_STATES.COMPLETED_DELETING_SERVERS) return True # server_start('hxckdwxwld-nested') # server_delete('oztfvquhhi-cybergym-publicprivate')
def server_start(server_name): """ Starts a server based on the specification in the Datastore entity with name server_name. A guacamole server is also registered with DNS. :param server_name: The Datastore entity name of the server to start :return: A boolean status on the success of the start """ server = ds_client.get(ds_client.key('cybergym-server', server_name)) state_transition(entity=server, new_state=SERVER_STATES.STARTING) workout_globals.refresh_api() response = compute.instances().start(project=project, zone=zone, instance=server_name).execute() print(f'Sent start request to {server_name}, and waiting for response') i = 0 success = False while not success and i < 5: try: print(f"Begin waiting for start response from operation {response['id']}") compute.zoneOperations().wait(project=project, zone=zone, operation=response["id"]).execute() success = True except timeout: i += 1 print('Response timeout for starting server. Trying again') pass if not success: print(f'Timeout in trying to start server {server_name}') state_transition(entity=server, new_state=SERVER_STATES.BROKEN) return False # If this is the guacamole server for student entry, then register the new DNS if 'student_entry' in server and server['student_entry']: print(f'Setting DNS record for {server_name}') ip_address = register_student_entry(server['workout'], server_name) server['external_ip'] = ip_address state_transition(entity=server, new_state=SERVER_STATES.RUNNING) print(f"Finished starting {server_name}") # If all servers have started, then change the build state build_id = server['workout'] check_build_state_change(build_id=build_id, check_server_state=SERVER_STATES.RUNNING, change_build_state=BUILD_STATES.RUNNING) return True
def build_student_servers(unit_id, workouts, student_entry_server, student_entry_type, student_entry_username, student_entry_password, network_type): """ Builds the student servers for the arena. In the arena, all student servers have the same configuration. However, they may be on distinct networks or the same network. :param unit_id: Unit ID for the Arena build. This is used to pull data from the datastore :param workouts: An array of workout_ids for this arena. :param student_entry_server: The name of the server used for students to access the arena :param student_entry_type: Either vnc or rdp :param student_entry_username: The username for the server or None for vnc :param student_entry_password: The password for the server :param network_type: Either a same network for all students or each student gets their distinct workout. :return: """ # STATE: BUILDING_ARENA_NETWORKS unit = ds_client.get(ds_client.key('cybergym-unit', unit_id)) if check_ordered_arenas_state(unit, BUILD_STATES.BUILDING_ARENA_STUDENT_NETWORKS): state_transition(entity=unit, new_state=BUILD_STATES.BUILDING_ARENA_STUDENT_NETWORKS) # If all of the student servers are in the same network, then create a single network first guac_network = '' if network_type == 'same': build_student_network(build_id=unit_id, subnet=student_network_subnet) network_name = "%s-%s" % (unit_id, student_network_name) subnet_name = "%s-%s-%s" % (unit_id, student_network_name, 'default') guac_network = network_name elif network_type == 'distinct': for workout_id in workouts: i = 0 if i == 0 else i + 1 build_student_network(build_id=workout_id, subnet='10.1.' + str(i) + '.0/24') network_name = "%s-%s" % (workout_id, student_network_name) subnet_name = "%s-%s-%s" % (workout_id, student_network_name, 'default') if not guac_network: guac_network = network_name state_transition(entity=unit, new_state=BUILD_STATES.COMPLETED_ARENA_STUDENT_NETWORKS) # STATE: BUILDING_ARENA_SERVERS # The IP address of each server is dynamically assigned as 10.1.i.j if check_ordered_arenas_state(unit, BUILD_STATES.BUILDING_ARENA_STUDENT_SERVERS): state_transition(entity=unit, new_state=BUILD_STATES.BUILDING_ARENA_STUDENT_SERVERS) i = 0 j = 2 guacamole_connections = [] for workout_id in workouts: workout_key = ds_client.key('cybergym-workout', workout_id) workout = ds_client.get(workout_key) for server in workout['student_servers']: internal_ip_address = f'10.1.{i}.{j}' if server['name'] == student_entry_server: guac_connection = { 'workout_id': workout_id, 'entry_type': student_entry_type, 'ip': internal_ip_address, 'username': student_entry_username, 'password': student_entry_password } guacamole_connections.append(guac_connection) server_name = "%s-%s" % (workout_id, server['name']) sshkey = server["sshkey"] tags = server['tags'] machine_type = server["machine_type"] network_routing = server["network_routing"] nics = [{ "network": network_name, "internal_IP": internal_ip_address, "subnet": subnet_name, "external_NAT": False }] create_instance_custom_image(compute=compute, workout=workout_id, name=server_name, custom_image=server['image'], machine_type=machine_type, networkRouting=network_routing, networks=nics, tags=tags, meta_data=None, sshkey=sshkey) j += 1 # Build the workout entry server and create the firewall rule to make it accessible. build_guacamole_server(build=unit, network=guac_network, guacamole_connections=guacamole_connections) # Get the unit datastore entry again because building the guacamole server adds credentials. unit = ds_client.get(ds_client.key('cybergym-unit', unit_id)) state_transition(entity=unit, new_state=BUILD_STATES.COMPLETED_ARENA_STUDENT_SERVERS)
def build_arena(unit_id): """ Builds an arena of student servers and a common compute environment according to the specification referenced in the workout_unit datastore :param unit_id: The workout_id key in the datastore holding the build specification :return: None """ key = ds_client.key('cybergym-unit', unit_id) unit = ds_client.get(key) # This can sometimes happen when debugging a Unit ID and the Datastore record no longer exists. arena = unit['arena'] if not arena: print('No unit %s exists in the data store' % unit_id) return if 'state' not in unit or not unit['state']: state_transition(entity=unit, new_state=BUILD_STATES.START) # # Parse the assessment specification to obtain any startup scripts for the workout. # startup_scripts = None # if unit['assessment']: # startup_scripts = get_startup_scripts(workout_id=workout_id, assessment=workout['assessment']) # # Create the networks and subnets # First create the student servers print('Creating student servers') build_student_servers(unit_id=unit_id, workouts=unit['workouts'], student_entry_type=arena['student_entry_type'], student_entry_server=arena['student_entry'], student_entry_username=arena['student_entry_username'], student_entry_password=arena['student_entry_password'], network_type=arena['student_network_type']) # STATE: BUILDING_ARENA_NETWORKS if check_ordered_arenas_state(unit, BUILD_STATES.BUILDING_ARENA_NETWORKS): state_transition(entity=unit, new_state=BUILD_STATES.BUILDING_ARENA_NETWORKS) if arena['networks']: print('Creating additional arena networks') create_network(networks=arena['networks'], build_id=unit_id) state_transition(entity=unit, new_state=BUILD_STATES.COMPLETED_ARENA_NETWORKS) # STATE: BUILDING_ARENA_SERVERS if check_ordered_arenas_state(unit, BUILD_STATES.BUILDING_ARENA_SERVERS): state_transition(entity=unit, new_state=BUILD_STATES.BUILDING_ARENA_SERVERS) print('Creating additional servers') i = 101 for server in arena['servers']: server_name = "%s-%s" % (unit_id, server['name']) sshkey = server["sshkey"] guac_path = server['guac_path'] tags = server['tags'] machine_type = server["machine_type"] network_routing = server["network_routing"] # If a nic is not specified, then add the server to the student-network. if server['nics']: nics = [] for n in server['nics']: if 'network' not in n: n['network'] = student_network_name if 'internal_IP' not in n: n['internal_IP'] = f'10.1.0.{i}' if 'subnet' not in n: n['subnet'] = 'default' if 'external_NAT' not in n: n['external_NAT'] = False nic = { "network": "%s-%s" % (unit_id, n['network']), "internal_IP": n['internal_IP'], "subnet": "%s-%s-%s" % (unit_id, n['network'], n['subnet']), "external_NAT": n['external_NAT'] } nics.append(nic) else: nics = [ { "network": "%s-%s" % (unit_id, student_network_name), "internal_IP": f'10.1.0.{i}', "subnet": "%s-%s-%s" % (unit_id, student_network_name, 'default'), "external_NAT": False } ] create_instance_custom_image(compute=compute, workout=unit_id, name=server_name, custom_image=server['image'], machine_type=machine_type, networkRouting=network_routing, networks=nics, tags=tags, meta_data=None, sshkey=sshkey) i += 1 state_transition(entity=unit, new_state=BUILD_STATES.COMPLETED_ARENA_SERVERS) # STATE: BUILDING_ROUTES if check_ordered_arenas_state(unit, BUILD_STATES.BUILDING_ROUTES): state_transition(entity=unit, new_state=BUILD_STATES.BUILDING_ROUTES) print('Creating network routes and firewall rules') if 'routes' in arena and arena['routes']: for route in arena['routes']: r = {"name": "%s-%s" % (unit_id, route['name']), "network": "%s-%s" % (unit_id, route['network']), "destRange": route['dest_range'], "nextHopInstance": "%s-%s" % (unit_id, route['next_hop_instance'])} create_route(route) state_transition(entity=unit, new_state=BUILD_STATES.COMPLETED_ROUTES) # STATE: BUILDING_FIREWALL if check_ordered_arenas_state(unit, BUILD_STATES.BUILDING_FIREWALL): state_transition(entity=unit, new_state=BUILD_STATES.BUILDING_FIREWALL) firewall_rules = [] for rule in arena['firewall_rules']: if 'network' not in rule: rule['network'] = student_network_name firewall_rules.append({"name": "%s-%s" % (unit_id, rule['name']), "network": "%s-%s" % (unit_id, rule['network']), "targetTags": rule['target_tags'], "protocol": rule['protocol'], "ports": rule['ports'], "sourceRanges": rule['source_ranges']}) # Create the default rules to allow traffic between student networks. firewall_rules.append({"name": "%s-%s" % (unit_id, 'allow-all-internal'), "network": "%s-%s" % (unit_id, student_network_name), "targetTags": [], "protocol": 'tcp', "ports": ['tcp/any', 'udp/any', 'icmp/any'], "sourceRanges": [student_network_subnet]}) create_firewall_rules(firewall_rules) state_transition(entity=unit, new_state=BUILD_STATES.COMPLETED_FIREWALL) state_transition(entity=unit, new_state=BUILD_STATES.READY)
def build_workout(workout_id): """ Builds a workout compute environment according to the specification referenced in the datastore with key workout_id :param workout_id: The workout_id key in the datastore holding the build specification :return: None """ key = ds_client.key('cybergym-workout', workout_id) workout = ds_client.get(key) # This can sometimes happen when debugging a workout ID and the Datastore record no longer exists. if not workout: cloud_log(workout_id, f"The datastore record for {workout_id} no longer exists!", LOG_LEVELS.ERROR) raise LookupError if 'state' not in workout or not workout['state']: state_transition(entity=workout, new_state=BUILD_STATES.START) # Create the networks and subnets if check_ordered_workout_state(workout, BUILD_STATES.BUILDING_NETWORKS): state_transition(entity=workout, new_state=BUILD_STATES.BUILDING_NETWORKS) for network in workout['networks']: cloud_log(workout_id, f"Building network {workout_id}-{network['name']}", LOG_LEVELS.INFO) network_body = { "name": f"{workout_id}-{network['name']}", "autoCreateSubnetworks": False, "region": region } try: response = compute.networks().insert( project=project, body=network_body).execute() compute.globalOperations().wait( project=project, operation=response["id"]).execute() time.sleep(10) except HttpError as err: # If the network already exists, then this may be a rebuild and ignore the error if err.resp.status in [409]: pass for subnet in network['subnets']: cloud_log( workout_id, f"Building the subnetwork {network_body['name']}-{subnet['name']}", LOG_LEVELS.INFO) subnetwork_body = { "name": f"{network_body['name']}-{subnet['name']}", "network": "projects/%s/global/networks/%s" % (project, network_body['name']), "ipCidrRange": subnet['ip_subnet'] } try: response = compute.subnetworks().insert( project=project, region=region, body=subnetwork_body).execute() compute.regionOperations().wait( project=project, region=region, operation=response["id"]).execute() except HttpError as err: # If the subnetwork already exists, then this may be a rebuild and ignore the error if err.resp.status in [409]: pass state_transition(entity=workout, new_state=BUILD_STATES.COMPLETED_NETWORKS) # Now create the server configurations if check_ordered_workout_state(workout, BUILD_STATES.BUILDING_SERVERS): state_transition(entity=workout, new_state=BUILD_STATES.BUILDING_SERVERS) pubsub_topic = PUBSUB_TOPICS.MANAGE_SERVER publisher = pubsub_v1.PublisherClient() topic_path = publisher.topic_path(project, pubsub_topic) for server in workout['servers']: server_name = f"{workout_id}-{server['name']}" cloud_log(workout_id, f"Sending pubsub message to build {server_name}", LOG_LEVELS.INFO) publisher.publish(topic_path, data=b'Server Build', server_name=server_name, action=SERVER_ACTIONS.BUILD) # Also build the student entry server for the workout publisher.publish(topic_path, data=b'Server Build', server_name=f"{workout_id}-student-guacamole", action=SERVER_ACTIONS.BUILD) state_transition(entity=workout, new_state=BUILD_STATES.COMPLETED_SERVERS) # Create all of the network routes and firewall rules if check_ordered_workout_state(workout, BUILD_STATES.BUILDING_ROUTES): state_transition(entity=workout, new_state=BUILD_STATES.BUILDING_ROUTES) cloud_log( workout_id, f"Creating network routes and firewall rules for {workout_id}", LOG_LEVELS.INFO) if 'routes' in workout and workout['routes']: workout_route_setup(workout_id) if check_ordered_workout_state(workout, BUILD_STATES.BUILDING_FIREWALL): state_transition(entity=workout, new_state=BUILD_STATES.BUILDING_FIREWALL) firewall_rules = [] for rule in workout['firewall_rules']: firewall_rules.append({ "name": "%s-%s" % (workout_id, rule['name']), "network": "%s-%s" % (workout_id, rule['network']), "targetTags": rule['target_tags'], "protocol": rule['protocol'], "ports": rule['ports'], "sourceRanges": rule['source_ranges'] }) create_firewall_rules(firewall_rules) state_transition(entity=workout, new_state=BUILD_STATES.COMPLETED_FIREWALL) cloud_log( workout_id, f"Finished the build process with a final state: {workout['state']}", LOG_LEVELS.INFO)
def server_build(server_name): """ Builds an individual server based on the specification in the Datastore entity with name server_name. :param server_name: The Datastore entity name of the server to build :return: A boolean status on the success of the build """ server = ds_client.get(ds_client.key('cybergym-server', server_name)) build_id = server['workout'] g_logger = log_client.logger(str(server_name)) state_transition(entity=server, new_state=SERVER_STATES.BUILDING) config = server['config'].copy() """ Currently, we need a workaround to insert the guacamole startup script because of a 1500 character limit on indexed fields. The exclude_from_index does not work on embedded datastore fields """ if 'student_entry' in server and server['student_entry']: config['metadata'] = { 'items': [{ "key": "startup-script", "value": server['guacamole_startup_script'] }] } # Begin the server build and keep trying for a bounded number of additional 30-second cycles i = 0 build_success = False while not build_success and i < 5: workout_globals.refresh_api() try: if server['add_disk']: try: image_config = { "name": server_name + "-disk", "sizeGb": server['add_disk'], "type": "projects/" + project + "/zones/" + zone + "/diskTypes/pd-ssd" } response = compute.disks().insert( project=project, zone=zone, body=image_config).execute() compute.zoneOperations().wait( project=project, zone=zone, operation=response["id"]).execute() except HttpError as err: # If the disk already exists (i.e. a nuke), then ignore if err.resp.status in [409]: pass if server['build_type'] == BUILD_TYPES.MACHINE_IMAGE: source_machine_image = f"projects/{project}/global/machineImages/{server['machine_image']}" compute_beta = discovery.build('compute', 'beta') response = compute_beta.instances().insert( project=project, zone=zone, body=config, sourceMachineImage=source_machine_image).execute() else: if "delayed_start" in server and server["delayed_start"]: time.sleep(30) response = compute.instances().insert(project=project, zone=zone, body=config).execute() build_success = True g_logger.log_text( f'Sent job to build {server_name}, and waiting for response') except BrokenPipeError: i += 1 except HttpError as exception: cloud_log( build_id, f"Error when trying to build {server_name}: {exception.reason}", LOG_LEVELS.ERROR) return False i = 0 success = False while not success and i < 5: try: g_logger.log_text( f"Begin waiting for build operation {response['id']}") compute.zoneOperations().wait(project=project, zone=zone, operation=response["id"]).execute() success = True except timeout: i += 1 g_logger.log_text('Response timeout for build. Trying again') pass if success: g_logger.log_text(f'Successfully built server {server_name}') state_transition(entity=server, new_state=SERVER_STATES.RUNNING, existing_state=SERVER_STATES.BUILDING) else: g_logger.log_text(f'Timeout in trying to build server {server_name}') state_transition(entity=server, new_state=SERVER_STATES.BROKEN) return False # If this is a student entry server, register the DNS if 'student_entry' in server and server['student_entry']: g_logger.log_text(f'Setting DNS record for {server_name}') ip_address = register_student_entry(server['workout'], server_name) server['external_ip'] = ip_address ds_client.put(server) server = ds_client.get(ds_client.key('cybergym-server', server_name)) # Now stop the server before completing g_logger.log_text(f'Stopping {server_name}') compute.instances().stop(project=project, zone=zone, instance=server_name).execute() state_transition(entity=server, new_state=SERVER_STATES.STOPPED) # If no other servers are building, then set the workout to the state of READY. check_build_state_change(build_id=build_id, check_server_state=SERVER_STATES.STOPPED, change_build_state=BUILD_STATES.READY)
def build(self): key = ds_client.key('cybergym-unit', self.unit_id) unit = ds_client.get(key) # This can sometimes happen when debugging a Unit ID and the Datastore record no longer exists. arena = unit['arena'] if not arena: cloud_log( self.unit_id, f"Build operation failed. No unit {self.unit_id} exists in the data store", LOG_LEVELS.ERROR) raise LookupError if 'state' not in unit or not unit['state']: state_transition(entity=unit, new_state=BUILD_STATES.START) # STATE: BUILDING_STUDENT_NETWORKS if check_ordered_arenas_state( unit, BUILD_STATES.BUILDING_ARENA_STUDENT_NETWORKS): cloud_log(self.unit_id, f"Creating student networks for arena {self.unit_id}", LOG_LEVELS.INFO) student_network = [{ 'name': self.STUDENT_NETWORK_NAME, 'subnets': [{ 'name': f'{self.unit_id}-{self.STUDENT_NETWORK_NAME}-default', 'ip_subnet': self.STUDENT_NETWORK_SUBNET }] }] try: create_network(networks=student_network, build_id=self.unit_id) except HttpError as err: if err.resp.status not in [409]: cloud_log( self.unit_id, f"Error when trying to create the student network for " f"the arena {self.unit_id}", LOG_LEVELS.ERROR) raise # STATE: BUILDING_ARENA_NETWORKS if check_ordered_arenas_state(unit, BUILD_STATES.BUILDING_ARENA_NETWORKS): state_transition(entity=unit, new_state=BUILD_STATES.BUILDING_ARENA_NETWORKS) if arena['networks']: cloud_log( self.unit_id, f"Creating additional arena networks for arena {self.unit_id}", LOG_LEVELS.INFO) create_network(networks=arena['networks'], build_id=self.unit_id) state_transition(entity=unit, new_state=BUILD_STATES.COMPLETED_ARENA_NETWORKS) # STATE: BUILDING_ARENA_SERVERS if check_ordered_arenas_state(unit, BUILD_STATES.BUILDING_ARENA_SERVERS): state_transition(entity=unit, new_state=BUILD_STATES.BUILDING_ARENA_SERVERS) cloud_log(self.unit_id, f"Creating additional servers for arena {self.unit_id}", LOG_LEVELS.INFO) server_list = list( ds_client.query(kind='cybergym-server').add_filter( 'workout', '=', self.unit_id).fetch()) for server in server_list: server_name = server['name'] cloud_log(self.unit_id, f"Sending pubsub message to build {server_name}", LOG_LEVELS.INFO) pubsub_topic = PUBSUB_TOPICS.MANAGE_SERVER publisher = pubsub_v1.PublisherClient() topic_path = publisher.topic_path(project, pubsub_topic) publisher.publish(topic_path, data=b'Server Build', server_name=server_name, action=SERVER_ACTIONS.BUILD) state_transition(entity=unit, new_state=BUILD_STATES.COMPLETED_ARENA_SERVERS) # STATE: BUILDING_ROUTES if check_ordered_arenas_state(unit, BUILD_STATES.BUILDING_ROUTES): state_transition(entity=unit, new_state=BUILD_STATES.BUILDING_ROUTES) cloud_log( self.unit_id, f"Creating network routes and firewall rules for arena {self.unit_id}", LOG_LEVELS.INFO) if 'routes' in arena and arena['routes']: for route in arena['routes']: r = { "name": f"{self.unit_id}-{route['name']}", "network": f"{self.unit_id}-{route['network']}", "destRange": route['dest_range'], "nextHopInstance": f"{self.unit_id}-{route['next_hop_instance']}" } create_route(route) state_transition(entity=unit, new_state=BUILD_STATES.COMPLETED_ROUTES) # STATE: BUILDING_FIREWALL if check_ordered_arenas_state(unit, BUILD_STATES.BUILDING_FIREWALL): state_transition(entity=unit, new_state=BUILD_STATES.BUILDING_FIREWALL) firewall_rules = [] for rule in arena['firewall_rules']: if 'network' not in rule: rule['network'] = self.STUDENT_NETWORK_NAME firewall_rules.append({ "name": f"{self.unit_id}-{rule['name']}", "network": f"{self.unit_id}-{rule['network']}", "targetTags": rule['target_tags'], "protocol": rule['protocol'], "ports": rule['ports'], "sourceRanges": rule['source_ranges'] }) # Create the default rules to allow traffic between student networks. firewall_rules.append({ "name": f"{self.unit_id}-allow-all-internal", "network": f"{self.unit_id}-{self.STUDENT_NETWORK_NAME}", "targetTags": [], "protocol": 'tcp', "ports": ['tcp/any', 'udp/any', 'icmp/any'], "sourceRanges": [self.STUDENT_NETWORK_SUBNET] }) create_firewall_rules(firewall_rules) state_transition(entity=unit, new_state=BUILD_STATES.COMPLETED_FIREWALL)
def server_delete(server_name): g_logger = log_client.logger(str(server_name)) server_list = list( ds_client.query(kind='cybergym-server').add_filter( 'name', '=', str(server_name)).fetch()) server_is_deleted = list( ds_client.query(kind='cybergym-server').add_filter( 'name', '=', str(server_name)).add_filter('state', '=', 'DELETED').fetch()) if server_is_deleted and server_list: g_logger.log_text(f'Server "' + server_name + '" has already been deleted.') return True elif not server_list: g_logger.log_text(f'Server of name "' + server_name + '" does not exist in datastore, unable to Delete.') return True else: server = ds_client.get(ds_client.key('cybergym-server', server_name)) state_transition(entity=server, new_state=SERVER_STATES.DELETING) # If there are snapshots associated with this server, then delete the snapshots. if 'snapshot' in server and server['snapshot']: Snapshot.delete_snapshot(server_name) workout_globals.refresh_api() try: response = compute.instances().delete(project=project, zone=zone, instance=server_name).execute() except HttpError as exception: # If the server is already deleted or no longer exists, state_transition(entity=server, new_state=SERVER_STATES.DELETED) g_logger.log_text(f"Finished deleting {server_name}") # If all servers in the workout have been deleted, then set the workout state to True build_id = server['workout'] check_build_state_change( build_id=build_id, check_server_state=SERVER_STATES.DELETED, change_build_state=BUILD_STATES.COMPLETED_DELETING_SERVERS) return True g_logger.log_text( f'Sent delete request to {server_name}, and waiting for response') i = 0 success = False while not success and i < 5: try: g_logger.log_text( f"Begin waiting for delete response from operation {response['id']}" ) compute.zoneOperations().wait(project=project, zone=zone, operation=response["id"]).execute() success = True except timeout: i += 1 g_logger.log_text( 'Response timeout for deleting server. Trying again') pass if not success: g_logger.log_text(f'Timeout in trying to delete server {server_name}') state_transition(entity=server, new_state=SERVER_STATES.BROKEN) return False # If this is a student entry server, delete the DNS if 'student_entry' in server and server['student_entry']: g_logger.log_text(f'Deleting DNS record for {server_name}') ip_address = server['external_ip'] delete_dns(server['workout'], ip_address) state_transition(entity=server, new_state=SERVER_STATES.DELETED) g_logger.log_text(f"Finished deleting {server_name}") # If all servers in the workout have been deleted, then set the workout state to True build_id = server['workout'] check_build_state_change( build_id=build_id, check_server_state=SERVER_STATES.DELETED, change_build_state=BUILD_STATES.COMPLETED_DELETING_SERVERS) return True
def build_workout(workout_id): """ Builds a workout compute environment according to the specification referenced in the datastore with key workout_id :param workout_id: The workout_id key in the datastore holding the build specification :return: None """ key = ds_client.key('cybergym-workout', workout_id) workout = ds_client.get(key) # This can sometimes happen when debugging a workout ID and the Datastore record no longer exists. if not workout: print('No workout for %s exists in the data store' % workout_id) return startup_scripts = None # Parse the assessment specification to obtain any startup scripts for the workout. if 'state' not in workout or not workout['state']: state_transition(entity=workout, new_state=BUILD_STATES.START) if workout['assessment']: startup_scripts = get_startup_scripts(workout_id=workout_id, assessment=workout['assessment']) # Create the networks and subnets if check_ordered_workout_state(workout, BUILD_STATES.BUILDING_NETWORKS): state_transition(entity=workout, new_state=BUILD_STATES.BUILDING_NETWORKS) print('Creating networks') for network in workout['networks']: network_body = { "name": "%s-%s" % (workout_id, network['name']), "autoCreateSubnetworks": False, "region": region } response = compute.networks().insert(project=project, body=network_body).execute() compute.globalOperations().wait( project=project, operation=response["id"]).execute() time.sleep(10) for subnet in network['subnets']: subnetwork_body = { "name": "%s-%s" % (network_body['name'], subnet['name']), "network": "projects/%s/global/networks/%s" % (project, network_body['name']), "ipCidrRange": subnet['ip_subnet'] } response = compute.subnetworks().insert( project=project, region=region, body=subnetwork_body).execute() compute.regionOperations().wait( project=project, region=region, operation=response["id"]).execute() state_transition(entity=workout, new_state=BUILD_STATES.COMPLETED_NETWORKS) # Now create the server configurations if check_ordered_workout_state(workout, BUILD_STATES.BUILDING_SERVERS): state_transition(entity=workout, new_state=BUILD_STATES.BUILDING_SERVERS) print('Creating servers') for server in workout['servers']: server_name = "%s-%s" % (workout_id, server['name']) sshkey = server["sshkey"] tags = server['tags'] machine_type = server["machine_type"] network_routing = server["network_routing"] min_cpu_platform = server[ "minCpuPlatform"] if "minCpuPlatform" in server else None nics = [] for n in server['nics']: nic = { "network": f"{workout_id}-{n['network']}", "internal_IP": n['internal_IP'], "subnet": f"{workout_id}-{n['network']}-{n['subnet']}", "external_NAT": n['external_NAT'] } # Nested VMs are sometimes used for vulnerable servers. This adds those specified IP addresses as # aliases to the NIC if 'IP_aliases' in n and n['IP_aliases']: alias_ip_ranges = [] for ipaddr in n['IP_aliases']: alias_ip_ranges.append({"ipCidrRange": ipaddr}) nic['aliasIpRanges'] = alias_ip_ranges nics.append(nic) # Add the startup script for assessment as metadata if it exists meta_data = None if startup_scripts and server['name'] in startup_scripts: meta_data = startup_scripts[server['name']] create_instance_custom_image(compute=compute, workout=workout_id, name=server_name, custom_image=server['image'], machine_type=machine_type, networkRouting=network_routing, networks=nics, tags=tags, meta_data=meta_data, sshkey=sshkey, minCpuPlatform=min_cpu_platform) state_transition(entity=workout, new_state=BUILD_STATES.COMPLETED_SERVERS) # Create the student entry guacamole server if check_ordered_workout_state(workout, BUILD_STATES.BUILDING_STUDENT_ENTRY): state_transition(entity=workout, new_state=BUILD_STATES.BUILDING_STUDENT_ENTRY) if workout['student_entry']: network_name = f"{workout_id}-{workout['student_entry']['network']}" student_entry_username = workout['student_entry'][ 'username'] if 'username' in workout['student_entry'] else None security_mode = workout['student_entry'][ 'security-mode'] if 'security-mode' in workout[ 'student_entry'] else 'nla' guac_connection = [{ 'workout_id': workout_id, 'entry_type': workout['student_entry']['type'], 'ip': workout['student_entry']['ip'], 'username': student_entry_username, 'password': workout['student_entry']['password'], 'security-mode': security_mode }] build_guacamole_server(build=workout, network=network_name, guacamole_connections=guac_connection) # Get the workout key again or the state transition will overwrite it workout = ds_client.get( ds_client.key('cybergym-workout', workout_id)) else: state_transition(entity=workout, new_state=BUILD_STATES.BROKEN) return state_transition(entity=workout, new_state=BUILD_STATES.COMPLETED_STUDENT_ENTRY) # Create all of the network routes and firewall rules if check_ordered_workout_state(workout, BUILD_STATES.BUILDING_ROUTES): state_transition(entity=workout, new_state=BUILD_STATES.BUILDING_ROUTES) print('Creating network routes and firewall rules') if 'routes' in workout and workout['routes']: for route in workout['routes']: response = compute.instances().get( project=project, zone=zone, instance=f"{workout_id}-{route['next_hop_instance']}") r = { "name": "%s-%s" % (workout_id, route['name']), "network": "%s-%s" % (workout_id, route['network']), "destRange": route['dest_range'], "nextHopInstance": "%s-%s" % (workout_id, route['next_hop_instance']) } create_route(r) if check_ordered_workout_state(workout, BUILD_STATES.BUILDING_FIREWALL): state_transition(entity=workout, new_state=BUILD_STATES.BUILDING_FIREWALL) firewall_rules = [] for rule in workout['firewall_rules']: firewall_rules.append({ "name": "%s-%s" % (workout_id, rule['name']), "network": "%s-%s" % (workout_id, rule['network']), "targetTags": rule['target_tags'], "protocol": rule['protocol'], "ports": rule['ports'], "sourceRanges": rule['source_ranges'] }) create_firewall_rules(firewall_rules) # build_workout('isirdhzjqk')
def server_build(server_name): """ Builds an individual server based on the specification in the Datastore entity with name server_name. :param server_name: The Datastore entity name of the server to build :return: A boolean status on the success of the build """ print(f'Building server {server_name}') server = ds_client.get(ds_client.key('cybergym-server', server_name)) state_transition(entity=server, new_state=SERVER_STATES.BUILDING) # Commented because this is only for Fortinet right now. # if 'canIPForward' in server and server['config']['canIpForward']: # image_config = {"name": server_name + "-disk", "sizeGb": 30, # "type": "projects/" + project + "/zones/" + zone + "/diskTypes/pd-ssd"} # response = compute.disks().insert(project=project, zone=zone, body=image_config).execute() # compute.zoneOperations().wait(project=project, zone=zone, operation=response["id"]).execute() # Begin the server build and keep trying for a bounded number of additional 30-second cycles i = 0 build_success = False while not build_success and i < 5: workout_globals.refresh_api() try: response = compute.instances().insert(project=project, zone=zone, body=server['config']).execute() build_success = True print(f'Sent job to build {server_name}, and waiting for response') except BrokenPipeError: i += 1 i = 0 success = False while not success and i < 5: try: print(f"Begin waiting for build operation {response['id']}") compute.zoneOperations().wait(project=project, zone=zone, operation=response["id"]).execute() success = True except timeout: i += 1 print('Response timeout for build. Trying again') pass if success: print(f'Successfully built server {server_name}') state_transition(entity=server, new_state=SERVER_STATES.RUNNING, existing_state=SERVER_STATES.BUILDING) else: print(f'Timeout in trying to build server {server_name}') state_transition(entity=server, new_state=SERVER_STATES.BROKEN) return False # If this is a student entry server, register the DNS if 'student_entry' in server and server['student_entry']: print(f'Setting DNS record for {server_name}') ip_address = register_student_entry(server['workout'], server_name) server['external_ip'] = ip_address ds_client.put(server) server = ds_client.get(ds_client.key('cybergym-server', server_name)) # Now stop the server before completing print(f'Stopping {server_name}') compute.instances().stop(project=project, zone=zone, instance=server_name).execute() state_transition(entity=server, new_state=SERVER_STATES.STOPPED) # If no other servers are building, then set the workout to the state of READY. build_id = server['workout'] check_build_state_change(build_id=build_id, check_server_state=SERVER_STATES.STOPPED, change_build_state=BUILD_STATES.READY)
def medic(): """ Reviews the state of all active workouts in the project and attempts to correct any which may have an invalid state. Invalid states often occur due to timeouts in processing the Google Cloud Functions. :returns: None """ g_logger = log_client.logger('workout-actions') g_logger.log_text("MEDIC: Running Medic function") # # Fixing build timeout issues # # The add_filter does not have a != operator. This provides an equivalent results for active workouts. query_current_workouts = ds_client.query(kind='cybergym-workout') results = list( query_current_workouts.add_filter('active', '=', True).fetch()) for workout in results: workout_project = workout.get('build_project_location', project) if workout_project == project: if get_workout_type(workout) == WORKOUT_TYPES.WORKOUT: if 'state' in workout: build_state = workout['state'] # if the workout state has not completed, then attempt to continue rebuilding the workout from where # it left off. if build_state in ordered_workout_build_states: g_logger.log_text( "MEDIC: Workout {} is in a build state of {}. Attempting to fix..." .format(workout.key.name, build_state)) build_workout(workout_id=workout.key.name) elif type(workout) is datastore.entity.Entity: # If there is no state, then this is not a valid workout, and we can delete the Datastore entity. g_logger.log_text( "Invalid workout specification in the datastore for workout ID: {}. Deleting the record." .format(workout.key.name)) ds_client.delete(workout.key) # # Fixing workouts in state COMPLETED_FIREWALL. This may occur when the firewall gets built after the guacamole server # query_completed_firewalls = ds_client.query(kind='cybergym-workout') results = list( query_completed_firewalls.add_filter( "state", "=", BUILD_STATES.COMPLETED_FIREWALL).fetch()) for workout in results: # Only transition the state if the last state change occurred over 5 minutes ago. workout_project = workout.get('build_project_location', project) if workout_project == project: if get_workout_type(workout) == WORKOUT_TYPES.WORKOUT: if workout['state-timestamp'] < str( calendar.timegm(time.gmtime()) - 300): g_logger.log_text( "MEDIC: Workout {} stuck in firewall completion. Changing state to READY" .format(workout.key.name)) state_transition(workout, new_state=BUILD_STATES.RUNNING) stop_workout(workout.key.name) # # Fixing workouts in state GUACAMOLE_SERVER_TIMEOUT. This may occur waiting for the guacamole server to come up # query_student_entry_timeouts = ds_client.query(kind='cybergym-workout') results = list( query_student_entry_timeouts.add_filter( "state", "=", BUILD_STATES.GUACAMOLE_SERVER_LOAD_TIMEOUT).fetch()) for workout in results: workout_project = workout.get('build_project_location', project) if workout_project == project: if get_workout_type(workout) == WORKOUT_TYPES.WORKOUT: # Change this to RUNNING unless the state change occurred over 15 minutes ago if workout['state-timestamp'] < str( calendar.timegm(time.gmtime()) - 900): g_logger.log_text( "MEDIC: Workout {} stuck in guacamole timeout. Changing state to READY" .format(workout.key.name)) state_transition(workout, new_state=BUILD_STATES.RUNNING) stop_workout(workout.key.name) else: g_logger.log_text( "MEDIC: Workout {} stuck in guacamole timeout. Changing state to READY" .format(workout.key.name)) print( f"Workout {workout.key.name} stuck in guacamole timeout. Changing state to READY" ) state_transition(workout, new_state=BUILD_STATES.READY) # # Fixing workouts in the state of STARTING. This may occur after a timeout in starting workouts. # query_start_timeouts = ds_client.query(kind='cybergym-workout') results = list( query_start_timeouts.add_filter("state", "=", BUILD_STATES.STARTING).fetch()) for workout in results: workout_project = workout.get('build_project_location', project) if workout_project == project: if get_workout_type(workout) == WORKOUT_TYPES.WORKOUT: # Only transition the state if the last state change occurred over 5 minutes ago. if workout['state-timestamp'] < str( calendar.timegm(time.gmtime()) - 300): g_logger.log_text( "MEDIC: Workout {} stuck in a STARTING state. Stopping the workout." .format(workout.key.name)) state_transition(workout, new_state=BUILD_STATES.RUNNING) stop_workout(workout.key.name) # # Fixing workouts in the state of STOPPING. This may occur after a timeout in stopping workouts. # query_stop_timeouts = ds_client.query(kind='cybergym-workout') results = list( query_stop_timeouts.add_filter("state", "=", BUILD_STATES.STOPPING).fetch()) for workout in results: workout_project = workout.get('build_project_location', project) if workout_project == project: if get_workout_type(workout) == WORKOUT_TYPES.WORKOUT: # Only transition the state if the last state change occurred over 5 minutes ago. if workout['state-timestamp'] < str( calendar.timegm(time.gmtime()) - 300): g_logger.log_text( "MEDIC: Workout {} stuck in a STARTING state. Stopping the workout." .format(workout.key.name)) state_transition(workout, new_state=BUILD_STATES.RUNNING) stop_workout(workout.key.name) # # Fixing workouts in the state of NUKING. This may occur after a timeout in deleting the workouts. # query_nuking_timeouts = ds_client.query(kind='cybergym-workout') results = list( query_nuking_timeouts.add_filter("state", "=", BUILD_STATES.NUKING).fetch()) for workout in results: workout_project = workout.get('build_project_location', project) if workout_project == project: if get_workout_type(workout) == WORKOUT_TYPES.WORKOUT: # Only transition the state if the last state change occurred over 5 minutes ago. if workout['state-timestamp'] < str( calendar.timegm(time.gmtime()) - 300): g_logger.log_text( "MEDIC: Workout {} stuck in a NUKING state. Attempting to nuke again." .format(workout.key.name)) nuke_workout(workout.key.name) # #Fixing machines that did not get built # query_rebuild = ds_client.query(kind='cybergym-workout') query_rebuild.add_filter('state', '=', BUILD_STATES.READY) query_rebuild.add_filter('build_project_location', '=', project) running_machines = list(query_rebuild.fetch()) current_machines = compute.instances().list(project=project, zone=zone).execute() list_current = [] list_running = [] list_missing = [] current_machines_items = current_machines.get('items', None) while current_machines_items: for instance in current_machines_items: list_current.append(instance['name']) if 'nextPageToken' in current_machines: current_machines = compute.instances().list( project=project, zone=zone, pageToken=current_machines['nextPageToken']).execute() current_machines_items = current_machines.get('items', None) else: break for i in running_machines: unit = ds_client.get(ds_client.key('cybergym-unit', i['unit_id'])) if unit['build_type'] == 'arena': for server in i['student_servers']: datastore_server_name = i.key.name + '-' + server['name'] list_running.append(datastore_server_name) if datastore_server_name not in list_current: list_missing.append(datastore_server_name) if unit['build_type'] == 'compute': for server in i['servers']: datastore_server_name = i.key.name + '-' + server['name'] list_running.append(datastore_server_name) if datastore_server_name not in list_current: list_missing.append(datastore_server_name) cloud_log('Medic', f'Missing servers{list_missing}', LOG_LEVELS.INFO) for server in list_missing: cloud_log('Medic', 'Rebuilding server{}'.format(server), LOG_LEVELS.INFO) pubsub_topic = PUBSUB_TOPICS.MANAGE_SERVER publisher = pubsub_v1.PublisherClient() topic_path = publisher.topic_path(project, pubsub_topic) future = publisher.publish(topic_path, data=b'Server Build', server_name=server, action=SERVER_ACTIONS.BUILD) return