| #copyright 2016-present Ciena Corporation |
| # |
| # Licensed under the Apache License, Version 2.0 (the "License"); |
| # you may not use this file except in compliance with the License. |
| # You may obtain a copy of the License at |
| # |
| # http://www.apache.org/licenses/LICENSE-2.0 |
| # |
| # Unless required by applicable law or agreed to in writing, software |
| # distributed under the License is distributed on an "AS IS" BASIS, |
| # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| # See the License for the specific language governing permissions and |
| # limitations under the License. |
| # |
| import unittest |
| from nose.tools import * |
| from scapy.all import * |
| from OnosCtrl import OnosCtrl, get_mac |
| from OltConfig import OltConfig |
| from socket import socket |
| from OnosFlowCtrl import OnosFlowCtrl |
| from nose.twistedtools import reactor, deferred |
| from twisted.internet import defer |
| from onosclidriver import OnosCliDriver |
| from CordContainer import Container, Onos, Quagga |
| from CordTestServer import cord_test_onos_restart, cord_test_onos_shutdown, cord_test_onos_add_cluster, cord_test_quagga_restart |
| from portmaps import g_subscriber_port_map |
| from scapy.all import * |
| import time, monotonic |
| import threading |
| from threading import current_thread |
| from Cluster import * |
| from EapTLS import TLSAuthTest |
| from ACL import ACLTest |
| import os |
| import json |
| import random |
| import collections |
| log.setLevel('INFO') |
| |
| class cluster_exchange(unittest.TestCase): |
| test_path = os.path.dirname(os.path.realpath(__file__)) |
| onos_config_path = os.path.join(test_path, '..', 'setup/onos-config') |
| mac = RandMAC()._fix() |
| flows_eth = Ether(src = RandMAC()._fix(), dst = RandMAC()._fix()) |
| igmp_eth = Ether(dst = '01:00:5e:00:00:16', type = ETH_P_IP) |
| igmp_ip = IP(dst = '224.0.0.22') |
| ONOS_INSTANCES = 3 |
| V_INF1 = 'veth0' |
| TLS_TIMEOUT = 100 |
| device_id = 'of:' + get_mac() |
| igmp = cluster_igmp() |
| igmp_groups = igmp.mcast_ip_range(start_ip = '224.1.8.10',end_ip = '224.1.10.49') |
| igmp_sources = igmp.source_ip_range(start_ip = '38.24.29.35',end_ip='38.24.35.56') |
| tls = cluster_tls() |
| flows = cluster_flows() |
| proxyarp = cluster_proxyarp() |
| vrouter = cluster_vrouter() |
| acl = cluster_acl() |
| dhcprelay = cluster_dhcprelay() |
| subscriber = cluster_subscriber() |
| |
| def get_controller(self): |
| controller = os.getenv('ONOS_CONTROLLER_IP') or 'localhost' |
| controller = controller.split(',')[0] |
| return controller |
| |
| def cliEnter(self,controller = None): |
| retries = 0 |
| while retries < 3: |
| self.cli = OnosCliDriver(controller = controller,connect = True) |
| if self.cli.handle: |
| break |
| else: |
| retries += 1 |
| time.sleep(2) |
| |
| def cliExit(self): |
| self.cli.disconnect() |
| |
| def verify_cluster_status(self,controller = None,onos_instances=ONOS_INSTANCES,verify=False): |
| tries = 0 |
| try: |
| self.cliEnter(controller = controller) |
| while tries <= 10: |
| cluster_summary = json.loads(self.cli.summary(jsonFormat = True)) |
| if cluster_summary: |
| log.info("cluster 'summary' command output is %s"%cluster_summary) |
| nodes = cluster_summary['nodes'] |
| if verify: |
| if nodes == onos_instances: |
| self.cliExit() |
| return True |
| else: |
| tries += 1 |
| time.sleep(1) |
| else: |
| if nodes >= onos_instances: |
| self.cliExit() |
| return True |
| else: |
| tries += 1 |
| time.sleep(1) |
| else: |
| tries += 1 |
| time.sleep(1) |
| self.cliExit() |
| return False |
| except: |
| raise |
| return False |
| |
| def get_cluster_current_member_ips(self,controller = None): |
| tries = 0 |
| cluster_ips = [] |
| try: |
| self.cliEnter(controller = controller) |
| while tries <= 10: |
| cluster_nodes = json.loads(self.cli.nodes(jsonFormat = True)) |
| if cluster_nodes: |
| log.info("cluster 'nodes' output is %s"%cluster_nodes) |
| cluster_ips = map(lambda c: c['id'], cluster_nodes) |
| self.cliExit() |
| cluster_ips.sort(lambda i1,i2: int(i1.split('.')[-1]) - int(i2.split('.')[-1])) |
| return cluster_ips |
| else: |
| tries += 1 |
| self.cliExit() |
| return cluster_ips |
| except: |
| raise Exception('Failed to get cluster members') |
| return cluster_ips |
| |
| def get_cluster_container_names_ips(self): |
| onos_names_ips = {} |
| onos_ips = self.get_cluster_current_member_ips() |
| onos_names_ips[onos_ips[0]] = Onos.NAME |
| for i in range(1, len(onos_ips)): |
| name = '{0}-{1}'.format(Onos.NAME,i+1) |
| onos_names_ips[onos_ips[i]] = name |
| |
| return onos_names_ips |
| |
| #identifying current master of a connected device, not tested |
| def get_cluster_current_master_standbys(self,controller=None,device_id=device_id): |
| master = None |
| standbys = [] |
| tries = 0 |
| try: |
| cli = self.cliEnter(controller = controller) |
| while tries <= 10: |
| roles = json.loads(self.cli.roles(jsonFormat = True)) |
| log.info("cluster 'roles' command output is %s"%roles) |
| if roles: |
| for device in roles: |
| log.info('Verifying device info in line %s'%device) |
| if device['id'] == device_id: |
| master = str(device['master']) |
| standbys = map(lambda d: str(d), device['standbys']) |
| log.info('Master and standbys for device %s are %s and %s'%(device_id, master, standbys)) |
| self.cliExit() |
| return master, standbys |
| self.cliExit() |
| return master, standbys |
| else: |
| tries += 1 |
| time.sleep(1) |
| self.cliExit() |
| return master, standbys |
| except: |
| raise Exception('Cannot get cluster master and standbys') |
| return master, standbys |
| |
| def change_master_current_cluster(self,new_master=None,device_id=device_id,controller=None): |
| if new_master is None: return False |
| self.cliEnter() |
| cmd = 'device-role' + ' ' + device_id + ' ' + new_master + ' ' + 'master' |
| command = self.cli.command(cmd = cmd, jsonFormat = False) |
| self.cliExit() |
| time.sleep(60) |
| master, standbys = self.get_cluster_current_master_standbys(controller=controller,device_id=device_id) |
| assert_equal(master,new_master) |
| log.info('Cluster master changed to %s successfully'%new_master) |
| |
| ############# Cluster Test cases ########################### |
| #pass |
| def test_onos_cluster_formation_verify(self,onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| log.info('Cluster exists with %d ONOS instances'%onos_instances) |
| |
| #nottest cluster not coming up properly if member goes down |
| def test_onos_cluster_adding_members(self,add = 2, onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| onos_ips = self.get_cluster_current_member_ips() |
| onos_instances = len(onos_ips)+add |
| log.info('Adding %d nodes to the ONOS cluster' %add) |
| cord_test_onos_add_cluster(count = add) |
| status = self.verify_cluster_status(onos_instances=onos_instances) |
| assert_equal(status, True) |
| |
| def test_onos_cluster_removing_master(self, onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(len(standbys),(onos_instances-1)) |
| onos_names_ips = self.get_cluster_container_names_ips() |
| master_onos_name = onos_names_ips[master] |
| log.info('Removing cluster current master %s'%(master)) |
| cord_test_onos_shutdown(node = master_onos_name) |
| time.sleep(60) |
| onos_instances -= 1 |
| status = self.verify_cluster_status(onos_instances = onos_instances,controller=standbys[0]) |
| assert_equal(status, True) |
| new_master, standbys = self.get_cluster_current_master_standbys(controller=standbys[0]) |
| assert_not_equal(master,new_master) |
| log.info('Successfully removed cluster master instance') |
| |
| def test_onos_cluster_removing_one_member(self, onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(len(standbys),(onos_instances-1)) |
| onos_names_ips = self.get_cluster_container_names_ips() |
| member_onos_name = onos_names_ips[standbys[0]] |
| log.info('Removing cluster member %s'%standbys[0]) |
| cord_test_onos_shutdown(node = member_onos_name) |
| time.sleep(60) |
| onos_instances -= 1 |
| status = self.verify_cluster_status(onos_instances = onos_instances,controller=master) |
| assert_equal(status, True) |
| |
| def test_onos_cluster_removing_two_members(self,onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(len(standbys),(onos_instances-1)) |
| onos_names_ips = self.get_cluster_container_names_ips() |
| member1_onos_name = onos_names_ips[standbys[0]] |
| member2_onos_name = onos_names_ips[standbys[1]] |
| log.info('Removing cluster member %s'%standbys[0]) |
| cord_test_onos_shutdown(node = member1_onos_name) |
| log.info('Removing cluster member %s'%standbys[1]) |
| cord_test_onos_shutdown(node = member2_onos_name) |
| time.sleep(60) |
| onos_instances = onos_instances - 2 |
| status = self.verify_cluster_status(onos_instances = onos_instances,controller=master) |
| assert_equal(status, True) |
| |
| def test_onos_cluster_removing_N_members(self,remove = 2, onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(len(standbys),(onos_instances-1)) |
| onos_names_ips = self.get_cluster_container_names_ips() |
| for i in range(remove): |
| member_onos_name = onos_names_ips[standbys[i]] |
| log.info('Removing onos container with name %s'%standbys[i]) |
| cord_test_onos_shutdown(node = member_onos_name) |
| time.sleep(60) |
| onos_instances = onos_instances - remove |
| status = self.verify_cluster_status(onos_instances = onos_instances, controller=master) |
| assert_equal(status, True) |
| |
| #nottest test cluster not coming up properly if member goes down |
| def test_onos_cluster_adding_and_removing_members(self,onos_instances = ONOS_INSTANCES ,add = 2, remove = 2): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| onos_ips = self.get_cluster_current_member_ips() |
| onos_instances = len(onos_ips)+add |
| log.info('Adding %d ONOS instances to the cluster'%add) |
| cord_test_onos_add_cluster(count = add) |
| status = self.verify_cluster_status(onos_instances=onos_instances) |
| assert_equal(status, True) |
| log.info('Removing %d ONOS instances from the cluster'%remove) |
| for i in range(remove): |
| name = '{}-{}'.format(Onos.NAME, onos_instances - i) |
| log.info('Removing onos container with name %s'%name) |
| cord_test_onos_shutdown(node = name) |
| time.sleep(60) |
| onos_instances = onos_instances-remove |
| status = self.verify_cluster_status(onos_instances=onos_instances) |
| assert_equal(status, True) |
| |
| #nottest cluster not coming up properly if member goes down |
| def test_onos_cluster_removing_and_adding_member(self,onos_instances = ONOS_INSTANCES,add = 1, remove = 1): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| onos_ips = self.get_cluster_current_member_ips() |
| onos_instances = onos_instances-remove |
| log.info('Removing %d ONOS instances from the cluster'%remove) |
| for i in range(remove): |
| name = '{}-{}'.format(Onos.NAME, len(onos_ips)-i) |
| log.info('Removing onos container with name %s'%name) |
| cord_test_onos_shutdown(node = name) |
| time.sleep(60) |
| status = self.verify_cluster_status(onos_instances=onos_instances) |
| assert_equal(status, True) |
| log.info('Adding %d ONOS instances to the cluster'%add) |
| cord_test_onos_add_cluster(count = add) |
| onos_instances = onos_instances+add |
| status = self.verify_cluster_status(onos_instances=onos_instances) |
| assert_equal(status, True) |
| |
| def test_onos_cluster_restart(self,onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| log.info('Restarting cluster') |
| cord_test_onos_restart() |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| |
| def test_onos_cluster_master_restart(self,onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| onos_names_ips = self.get_cluster_container_names_ips() |
| master_onos_name = onos_names_ips[master] |
| log.info('Restarting cluster master %s'%master) |
| cord_test_onos_restart(node = master_onos_name) |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| log.info('Cluster came up after master restart as expected') |
| |
| #test fail. master changing after restart. Need to check correct behavior. |
| def test_onos_cluster_master_ip_after_master_restart(self,onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| master1, standbys = self.get_cluster_current_master_standbys() |
| onos_names_ips = self.get_cluster_container_names_ips() |
| master_onos_name = onos_names_ips[master1] |
| log.info('Restarting cluster master %s'%master) |
| cord_test_onos_restart(node = master_onos_name) |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| master2, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(master1,master2) |
| log.info('Cluster master is same before and after cluster master restart as expected') |
| |
| def test_onos_cluster_one_member_restart(self,onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(len(standbys),(onos_instances-1)) |
| onos_names_ips = self.get_cluster_container_names_ips() |
| member_onos_name = onos_names_ips[standbys[0]] |
| log.info('Restarting cluster member %s'%standbys[0]) |
| cord_test_onos_restart(node = member_onos_name) |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| log.info('Cluster came up as expected after restarting one member') |
| |
| def test_onos_cluster_two_members_restart(self,onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(len(standbys),(onos_instances-1)) |
| onos_names_ips = self.get_cluster_container_names_ips() |
| member1_onos_name = onos_names_ips[standbys[0]] |
| member2_onos_name = onos_names_ips[standbys[1]] |
| log.info('Restarting cluster members %s and %s'%(standbys[0],standbys[1])) |
| cord_test_onos_restart(node = member1_onos_name) |
| cord_test_onos_restart(node = member2_onos_name) |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| log.info('Cluster came up as expected after restarting two members') |
| |
| def test_onos_cluster_N_members_restart(self, members = 2, onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status,True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(len(standbys),(onos_instances-1)) |
| onos_names_ips = self.get_cluster_container_names_ips() |
| for i in range(members): |
| member_onos_name = onos_names_ips[standbys[i]] |
| log.info('Restarting cluster member %s'%standbys[i]) |
| cord_test_onos_restart(node = member_onos_name) |
| |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| log.info('Cluster came up as expected after restarting %d members'%members) |
| |
| def test_onos_cluster_master_change(self,onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances=onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(len(standbys),(onos_instances-1)) |
| log.info('Cluster current master of device is %s'%master) |
| self.change_master_current_cluster(new_master=standbys[0]) |
| log.info('Cluster master changed successfully') |
| |
| #tested on single onos setup. |
| def test_onos_cluster_vrouter_routes_in_cluster_members(self,networks = 5,onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| onos_ips = self.get_cluster_current_member_ips() |
| self.vrouter.setUpClass() |
| res = self.vrouter.vrouter_network_verify(networks, peers = 1) |
| assert_equal(res, True) |
| for onos_ip in onos_ips: |
| tries = 0 |
| flag = False |
| try: |
| self.cliEnter(controller = onos_ip) |
| while tries <= 5: |
| routes = json.loads(self.cli.routes(jsonFormat = True)) |
| if routes: |
| assert_equal(len(routes['routes4']), networks) |
| self.cliExit() |
| flag = True |
| break |
| else: |
| tries += 1 |
| time.sleep(1) |
| assert_equal(flag, True) |
| except: |
| log.info('Exception occured while checking routes in onos instance %s'%onos_ip) |
| raise |
| |
| #tested on single onos setup. |
| def test_onos_cluster_vrouter_master_down(self,networks = 5, onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| onos_ips = self.get_cluster_current_member_ips() |
| master, standbys = self.get_cluster_current_master_standbys() |
| onos_names_ips = self.get_cluster_container_names_ips() |
| master_onos_name = onos_names_ips[master] |
| self.vrouter.setUpClass() |
| res = self.vrouter.vrouter_network_verify(networks, peers = 1) |
| assert_equal(res,True) |
| cord_test_onos_shutdown(node = master_onos_name) |
| time.sleep(60) |
| log.info('Verifying vrouter traffic after cluster master down') |
| self.vrouter.vrouter_traffic_verify() |
| |
| #tested on single onos setup. |
| def test_onos_cluster_with_vrouter_and_restarting_master(self,networks = 5,onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| onos_ips = self.get_cluster_current_member_ips() |
| master, standbys = self.get_cluster_current_master_standbys() |
| onos_names_ips = self.get_cluster_container_names_ips() |
| master_onos_name = onos_names_ips[master] |
| self.vrouter.setUpClass() |
| res = self.vrouter.vrouter_network_verify(networks, peers = 1) |
| assert_equal(res, True) |
| cord_test_onos_restart() |
| self.vrouter.vrouter_traffic_verify() |
| |
| #tested on single onos setup. |
| def test_onos_cluster_deactivating_vrouter_app(self,networks = 5, onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| self.vrouter.setUpClass() |
| res = self.vrouter.vrouter_network_verify(networks, peers = 1) |
| assert_equal(res, True) |
| self.vrouter.vrouter_activate(deactivate=True) |
| time.sleep(15) |
| self.vrouter.vrouter_traffic_verify(positive_test=False) |
| self.vrouter.vrouter_activate(deactivate=False) |
| |
| #tested on single onos setup. |
| def test_onos_cluster_deactivating_vrouter_app_and_making_master_down(self,networks = 5,onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| onos_names_ips = self.get_cluster_container_names_ips() |
| master_onos_name = onos_names_ips[master] |
| self.vrouter.setUpClass() |
| log.info('Verifying vrouter before master down') |
| res = self.vrouter.vrouter_network_verify(networks, peers = 1) |
| assert_equal(res, True) |
| self.vrouter.vrouter_activate(deactivate=True) |
| log.info('Verifying vrouter traffic after app deactivated') |
| time.sleep(15) ## Expecting vrouter should work properly if master of cluster goes down |
| self.vrouter.vrouter_traffic_verify(positive_test=False) |
| log.info('Verifying vrouter traffic after master down') |
| cord_test_onos_shutdown(node = master_onos_name) |
| time.sleep(60) |
| self.vrouter.vrouter_traffic_verify(positive_test=False) |
| self.vrouter.vrouter_activate(deactivate=False) |
| |
| #tested on single onos setup. |
| def test_onos_cluster_for_vrouter_app_and_making_member_down(self,networks = 5,onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| onos_names_ips = self.get_cluster_container_names_ips() |
| member_onos_name = onos_names_ips[standbys[0]] |
| self.vrouter.setUpClass() |
| log.info('Verifying vrouter before cluster member down') |
| res = self.vrouter.vrouter_network_verify(networks, peers = 1) |
| assert_equal(res, True) # Expecting vrouter should work properly |
| log.info('Verifying vrouter after cluster member down') |
| cord_test_onos_shutdown(node = member_onos_name) |
| time.sleep(60) |
| self.vrouter.vrouter_traffic_verify()# Expecting vrouter should work properly if member of cluster goes down |
| |
| #tested on single onos setup. |
| def test_onos_cluster_for_vrouter_app_and_restarting_member(self,networks = 5, onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| onos_names_ips = self.get_cluster_container_names_ips() |
| member_onos_name = onos_names_ips[standbys[1]] |
| self.vrouter.setUpClass() |
| log.info('Verifying vrouter traffic before cluster member restart') |
| res = self.vrouter.vrouter_network_verify(networks, peers = 1) |
| assert_equal(res, True) # Expecting vrouter should work properly |
| cord_test_onos_restart(node = member_onos_name) |
| log.info('Verifying vrouter traffic after cluster member restart') |
| self.vrouter.vrouter_traffic_verify()# Expecting vrouter should work properly if member of cluster restarts |
| |
| #tested on single onos setup. |
| def test_onos_cluster_for_vrouter_app_restarting_cluster(self,networks = 5, onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| self.vrouter.setUpClass() |
| log.info('Verifying vrouter traffic before cluster restart') |
| res = self.vrouter.vrouter_network_verify(networks, peers = 1) |
| assert_equal(res, True) # Expecting vrouter should work properly |
| cord_test_onos_restart() |
| log.info('Verifying vrouter traffic after cluster restart') |
| self.vrouter.vrouter_traffic_verify()# Expecting vrouter should work properly if member of cluster restarts |
| |
| |
| #test fails because flow state is in pending_add in onos |
| def test_onos_cluster_for_flows_of_udp_port_and_making_master_down(self, onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances = onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| onos_names_ips = self.get_cluster_container_names_ips() |
| master_onos_name = onos_names_ips[master] |
| self.flows.setUpClass() |
| egress = 1 |
| ingress = 2 |
| egress_map = { 'ip': '192.168.30.1', 'udp_port': 9500 } |
| ingress_map = { 'ip': '192.168.40.1', 'udp_port': 9000 } |
| flow = OnosFlowCtrl(deviceId = self.device_id, |
| egressPort = egress, |
| ingressPort = ingress, |
| udpSrc = ingress_map['udp_port'], |
| udpDst = egress_map['udp_port'], |
| controller=master |
| ) |
| result = flow.addFlow() |
| assert_equal(result, True) |
| time.sleep(1) |
| self.success = False |
| def mac_recv_task(): |
| def recv_cb(pkt): |
| log.info('Pkt seen with ingress UDP port %s, egress UDP port %s' %(pkt[UDP].sport, pkt[UDP].dport)) |
| self.success = True |
| sniff(timeout=2, |
| lfilter = lambda p: UDP in p and p[UDP].dport == egress_map['udp_port'] |
| and p[UDP].sport == ingress_map['udp_port'], prn = recv_cb, iface = self.flows.port_map[egress]) |
| |
| for i in [0,1]: |
| if i == 1: |
| cord_test_onos_shutdown(node = master_onos_name) |
| log.info('Verifying flows traffic after master killed') |
| time.sleep(45) |
| else: |
| log.info('Verifying flows traffic before master killed') |
| t = threading.Thread(target = mac_recv_task) |
| t.start() |
| L2 = self.flows_eth #Ether(src = ingress_map['ether'], dst = egress_map['ether']) |
| L3 = IP(src = ingress_map['ip'], dst = egress_map['ip']) |
| L4 = UDP(sport = ingress_map['udp_port'], dport = egress_map['udp_port']) |
| pkt = L2/L3/L4 |
| log.info('Sending packets to verify if flows are correct') |
| sendp(pkt, count=50, iface = self.flows.port_map[ingress]) |
| t.join() |
| assert_equal(self.success, True) |
| |
| def test_onos_cluster_making_master_change_and_flows_of_ecn(self,onos_instances = ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances=onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| self.flows.setUpClass() |
| egress = 1 |
| ingress = 2 |
| egress_map = { 'ip': '192.168.30.1' } |
| ingress_map = { 'ip': '192.168.40.1' } |
| flow = OnosFlowCtrl(deviceId = self.device_id, |
| egressPort = egress, |
| ingressPort = ingress, |
| ecn = 1, |
| controller=master |
| ) |
| result = flow.addFlow() |
| assert_equal(result, True) |
| ##wait for flows to be added to ONOS |
| time.sleep(1) |
| self.success = False |
| def mac_recv_task(): |
| def recv_cb(pkt): |
| log.info('Pkt seen with ingress ip %s, egress ip %s and Type of Service %s' %(pkt[IP].src, pkt[IP].dst, pkt[IP].tos)) |
| self.success = True |
| sniff(count=2, timeout=5, |
| lfilter = lambda p: IP in p and p[IP].dst == egress_map['ip'] and p[IP].src == ingress_map['ip'] |
| and int(bin(p[IP].tos).split('b')[1][-2:],2) == 1,prn = recv_cb, |
| iface = self.flows.port_map[egress]) |
| for i in [0,1]: |
| if i == 1: |
| log.info('Changing cluster master to %s'%standbys[0]) |
| self.change_master_current_cluster(new_master=standbys[0]) |
| log.info('Verifying flow traffic after cluster master chnaged') |
| else: |
| log.info('Verifying flow traffic before cluster master changed') |
| t = threading.Thread(target = mac_recv_task) |
| t.start() |
| L2 = self.flows_eth # Ether(src = ingress_map['ether'], dst = egress_map['ether']) |
| L3 = IP(src = ingress_map['ip'], dst = egress_map['ip'], tos = 1) |
| pkt = L2/L3 |
| log.info('Sending a packet to verify if flows are correct') |
| sendp(pkt, count=50, iface = self.flows.port_map[ingress]) |
| t.join() |
| assert_equal(self.success, True) |
| |
| @deferred(TLS_TIMEOUT) |
| def test_onos_cluster_with_eap_tls_traffic(self,onos_instances=ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances=onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(len(standbys), (onos_instances-1)) |
| self.tls.setUp(controller=master) |
| df = defer.Deferred() |
| def eap_tls_verify(df): |
| tls = TLSAuthTest() |
| tls.runTest() |
| df.callback(0) |
| reactor.callLater(0, eap_tls_verify, df) |
| return df |
| |
| @deferred(120) |
| def test_onos_cluster_for_eap_tls_traffic_before_and_after_master_change(self,onos_instances=ONOS_INSTANCES): |
| master, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(len(standbys), (onos_instances-1)) |
| self.tls.setUp() |
| df = defer.Deferred() |
| def eap_tls_verify2(df2): |
| tls = TLSAuthTest() |
| tls.runTest() |
| df.callback(0) |
| for i in [0,1]: |
| if i == 1: |
| log.info('Changing cluster master %s to %s'%(master, standbys[0])) |
| self.change_master_current_cluster(new_master=standbys[0]) |
| log.info('Verifying tls authentication after cluster master changed to %s'%standbys[0]) |
| else: |
| log.info('Verifying tls authentication before cluster master change') |
| reactor.callLater(0, eap_tls_verify, df) |
| return df |
| |
| @deferred(TLS_TIMEOUT) |
| def test_onos_cluster_for_eap_tls_traffic_before_and_after_making_master_down(self,onos_instances=ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances=onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(len(standbys), (onos_instances-1)) |
| onos_names_ips = self.get_cluster_container_names_ips() |
| master_onos_name = onos_names_ips[master] |
| self.tls.setUp() |
| df = defer.Deferred() |
| def eap_tls_verify(df): |
| tls = TLSAuthTest() |
| tls.runTest() |
| df.callback(0) |
| for i in [0,1]: |
| if i == 1: |
| log.info('Killing cluster current master %s'%master) |
| cord_test_onos_shutdown(node = master_onos_name) |
| time.sleep(20) |
| status = self.verify_cluster_status(controller=standbys[0],onos_instances=onos_instances-1,verify=True) |
| assert_equal(status, True) |
| log.info('Cluster came up with %d instances after killing master'%(onos_instances-1)) |
| log.info('Verifying tls authentication after killing cluster master') |
| reactor.callLater(0, eap_tls_verify, df) |
| return df |
| |
| @deferred(TLS_TIMEOUT) |
| def test_onos_cluster_for_eap_tls_with_no_cert_before_and_after_member_is_restarted(self,onos_instances=ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances=onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(len(standbys), (onos_instances-1)) |
| onos_names_ips = self.get_cluster_container_names_ips() |
| member_onos_name = onos_names_ips[standbys[0]] |
| self.tls.setUp() |
| df = defer.Deferred() |
| def eap_tls_no_cert(df): |
| def tls_no_cert_cb(): |
| log.info('TLS authentication failed with no certificate') |
| tls = TLSAuthTest(fail_cb = tls_no_cert_cb, client_cert = '') |
| tls.runTest() |
| assert_equal(tls.failTest, True) |
| df.callback(0) |
| for i in [0,1]: |
| if i == 1: |
| log.info('Restart cluster member %s'%standbys[0]) |
| Container(member_onos_name,Onos.IMAGE).restart() |
| time.sleep(20) |
| status = self.verify_cluster_status(onos_instances=onos_instances) |
| assert_equal(status, True) |
| log.info('Cluster came up with %d instances after member restart'%(onos_instances)) |
| log.info('Verifying tls authentication after member restart') |
| reactor.callLater(0, eap_tls_no_cert, df) |
| return df |
| |
| ###### Dhcp Relay Test cases ###################################### |
| |
| def test_onos_cluster_with_dhcpRelay_app_releasing_dhcp_ip_after_master_change(self, iface = 'veth0',onos_instances=ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances=onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(len(standbys),(onos_instances-1)) |
| self.dhcprelay.setUpClass() |
| mac = self.dhcprelay.get_mac(iface) |
| self.dhcprelay.host_load(iface) |
| ##we use the defaults for this test that serves as an example for others |
| ##You don't need to restart dhcpd server if retaining default config |
| config = self.dhcprelay.default_config |
| options = self.dhcprelay.default_options |
| subnet = self.dhcprelay.default_subnet_config |
| dhcpd_interface_list = self.dhcprelay.relay_interfaces |
| self.dhcprelay.dhcpd_start(intf_list = dhcpd_interface_list, |
| config = config, |
| options = options, |
| subnet = subnet) |
| self.dhcprelay.dhcp = DHCPTest(seed_ip = '10.10.100.10', iface = iface) |
| cip, sip = self.dhcprelay.send_recv(mac) |
| log.info('Changing cluster current master from %s to %s'%(master, standbys[0])) |
| self.change_master_current_cluster(new_master=standbys[0]) |
| log.info('Releasing ip %s to server %s' %(cip, sip)) |
| assert_equal(self.dhcprelay.dhcp.release(cip), True) |
| log.info('Triggering DHCP discover again after release') |
| cip2, sip2 = self.dhcprelay.send_recv(mac) |
| log.info('Verifying released IP was given back on rediscover') |
| assert_equal(cip, cip2) |
| log.info('Test done. Releasing ip %s to server %s' %(cip2, sip2)) |
| assert_equal(self.dhcprelay.dhcp.release(cip2), True) |
| self.dhcprelay.tearDownClass() |
| |
| def test_onos_cluster_with_dhcpRelay_app_simulating_client_by_changing_master(self, iface = 'veth0',onos_instances=ONOS_INSTANCES): |
| #status = self.verify_cluster_status(onos_instances=onos_instances) |
| #assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(len(standbys),(onos_instances-1)) |
| self.dhcprelay.setUpClass() |
| macs = ['e4:90:5e:a3:82:c1','e4:90:5e:a3:82:c2','e4:90:5e:a3:82:c3'] |
| self.dhcprelay.host_load(iface) |
| ##we use the defaults for this test that serves as an example for others |
| ##You don't need to restart dhcpd server if retaining default config |
| config = self.dhcprelay.default_config |
| options = self.dhcprelay.default_options |
| subnet = self.dhcprelay.default_subnet_config |
| dhcpd_interface_list = self.dhcprelay.relay_interfaces |
| self.dhcprelay.dhcpd_start(intf_list = dhcpd_interface_list, |
| config = config, |
| options = options, |
| subnet = subnet) |
| self.dhcprelay.dhcp = DHCPTest(seed_ip = '10.10.10.1', iface = iface) |
| cip1, sip1 = self.dhcprelay.send_recv(macs[0]) |
| assert_not_equal(cip1,None) |
| log.info('Got dhcp client IP %s for mac %s when cluster master is %s'%(cip1,macs[0],master)) |
| log.info('Changing cluster master from %s to %s'%(master, standbys[0])) |
| self.change_master_current_cluster(new_master=standbys[0]) |
| cip2, sip2 = self.dhcprelay.send_recv(macs[1]) |
| assert_not_equal(cip2,None) |
| log.info('Got dhcp client IP %s for mac %s when cluster master is %s'%(cip2,macs[1],standbys[0])) |
| self.change_master_current_cluster(new_master=master) |
| log.info('Changing cluster master from %s to %s'%(standbys[0],master)) |
| cip3, sip3 = self.dhcprelay.send_recv(macs[2]) |
| assert_not_equal(cip3,None) |
| log.info('Got dhcp client IP %s for mac %s when cluster master is %s'%(cip2,macs[2],master)) |
| self.dhcprelay.tearDownClass() |
| |
| |
| ############ Cord Subscriber Test cases ################## |
| |
| def test_onos_cluster_with_cord_subscriber_joining_next_channel_before_and_after_cluster_restart(self,onos_instances=ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances=onos_instances) |
| assert_equal(status, True) |
| """Test subscriber join next for channel surfing""" |
| self.subscriber.setUpClass() |
| self.subscriber.num_subscribers = 5 |
| self.subscriber.num_channels = 10 |
| for i in [0,1]: |
| if i == 1: |
| cord_test_onos_restart() |
| time.sleep(45) |
| status = self.verify_cluster_status(onos_instances=onos_instances) |
| assert_equal(status, True) |
| log.info('Verifying cord subscriber functionality after cluster restart') |
| else: |
| log.info('Verifying cord subscriber functionality before cluster restart') |
| test_status = self.subscriber.subscriber_join_verify(num_subscribers = self.subscriber.num_subscribers, |
| num_channels = self.subscriber.num_channels, |
| cbs = (self.subscriber.tls_verify, self.subscriber.dhcp_next_verify, |
| self.subscriber.igmp_next_verify, self.subscriber.traffic_verify), |
| port_list = self.subscriber.generate_port_list(self.subscriber.num_subscribers, |
| self.subscriber.num_channels)) |
| assert_equal(test_status, True) |
| self.subscriber.tearDownClass() |
| |
| def test_onos_cluster_with_cord_subscriber_joining_10channels_making_one_cluster_member_down(self,onos_instances=ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances=onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(len(standbys),(onos_instances-1)) |
| onos_names_ips = self.get_cluster_container_names_ips() |
| member_onos_name = onos_names_ips[standbys[0]] |
| self.subscriber.setUpClass() |
| num_subscribers = 1 |
| num_channels = 10 |
| for i in [0,1]: |
| if i == 1: |
| cord_test_onos_shutdown(node = member_onos_name) |
| time.sleep(30) |
| status = self.verify_cluster_status(onos_instances=onos_instances-1,verify=True) |
| assert_equal(status, True) |
| log.info('Verifying cord subscriber functionality after cluster member %s is down'%standbys[0]) |
| else: |
| log.info('Verifying cord subscriber functionality before cluster member %s is down'%standbys[0]) |
| test_status = self.subscriber.subscriber_join_verify(num_subscribers = num_subscribers, |
| num_channels = num_channels, |
| cbs = (self.subscriber.tls_verify, self.subscriber.dhcp_verify, |
| self.subscriber.igmp_verify, self.subscriber.traffic_verify), |
| port_list = self.subscriber.generate_port_list(num_subscribers, num_channels), |
| negative_subscriber_auth = 'all') |
| assert_equal(test_status, True) |
| self.subscriber.tearDownClass() |
| |
| def test_onos_cluster_cord_subscriber_joining_next_10channels_making_two_cluster_members_down(self,onos_instances=ONOS_INSTANCES): |
| status = self.verify_cluster_status(onos_instances=onos_instances) |
| assert_equal(status, True) |
| master, standbys = self.get_cluster_current_master_standbys() |
| assert_equal(len(standbys),(onos_instances-1)) |
| onos_names_ips = self.get_cluster_container_names_ips() |
| member1_onos_name = onos_names_ips[standbys[0]] |
| member2_onos_name = onos_names_ips[standbys[1]] |
| self.subscriber.setUpClass() |
| num_subscribers = 1 |
| num_channels = 10 |
| for i in [0,1]: |
| if i == 1: |
| cord_test_onos_shutdown(node = member1_onos_name) |
| cord_test_onos_shutdown(node = member2_onos_name) |
| time.sleep(60) |
| status = self.verify_cluster_status(onos_instances=onos_instances-2) |
| assert_equal(status, True) |
| log.info('Verifying cord subscriber funtionality after cluster two members %s and %s down'%(standbys[0],standbys[1])) |
| else: |
| log.info('Verifying cord subscriber funtionality before cluster two members %s and %s down'%(standbys[0],standbys[1])) |
| test_status = self.subscriber.subscriber_join_verify(num_subscribers = num_subscribers, |
| num_channels = num_channels, |
| cbs = (self.subscriber.tls_verify, self.subscriber.dhcp_next_verify, |
| self.subscriber.igmp_next_verify, self.subscriber.traffic_verify), |
| port_list = self.subscriber.generate_port_list(num_subscribers, num_channels), |
| negative_subscriber_auth = 'all') |
| assert_equal(test_status, True) |
| self.subscriber.tearDownClass() |
| |