def cleanup_iptables(self):
while True:
if util.iptables_redirect('D', '127.0.0.100', '127.0.0.1') == False:
break
while True:
if util.iptables_drop('D', '127.0.0.100') == False:
break
while True:
if util.iptables_drop('D', '127.0.0.101') == False:
break
while True:
if util.iptables_drop('D', '127.0.0.102') == False:
break
def unblock_network(cluster, mgmt_ip, mgmt_port, final_state):
# Unblock
if util.iptables_drop('D', '127.0.0.100') == False:
util.log('delete a bloking role to iptables fail.')
return False
# Check cluster state
for i in range(3):
util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, final_state)
time.sleep(1)
return True
def block_network(cluster, mgmt_ip, mgmt_port):
# Block
if util.iptables_drop('A', '127.0.0.100') == False:
util.log('add a bloking role to iptables fail.')
return False
for i in range(4):
util.log('waiting... %d' % (i + 1))
time.sleep(1)
# Check cluster state
for i in range(2):
util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port)
time.sleep(1)
return True
def test_1_mgmt_is_isolated(self):
util.print_frame()
util.iptables_print_list()
cluster = filter(lambda x: x['cluster_name'] == 'network_isolation_cluster_1', config.clusters)[0]
util.log(util.json_to_str(cluster))
# MGMT
mgmt_ip = cluster['servers'][0]['real_ip']
mgmt_port = cluster['servers'][0]['cm_port']
# Create cluster
conf_checker = default_cluster.initialize_starting_up_smr_before_redis( cluster )
self.assertIsNotNone(conf_checker, 'failed to initialize cluster')
# Print initial state of cluster
util.log('\n\n\n ### INITIAL STATE OF CLUSTER ### ')
initial_state = []
self.assertTrue(util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, initial_state, check_quorum=True), 'failed to check cluster state')
# Set SMR option (slave_idle_timeout)
util.log('\n\n\n ### Set SMR option ###')
for s in cluster['servers']:
t = telnet.Telnet('SMR%d' % s['id'])
self.assertEqual(t.connect(s['ip'], s['smr_mgmt_port']), 0,
'Failed to connect to smr. ADDR=%s:%d' % (s['ip'], s['smr_mgmt_port']))
cmd = 'confset slave_idle_timeout_msec 18000'
util.log('[%s:%d] >> %s' % (s['ip'], s['smr_mgmt_port'], cmd))
t.write('confset slave_idle_timeout_msec 18000\r\n')
reply = t.read_until('\r\n').strip()
util.log('[%s:%d] << %s' % (s['ip'], s['smr_mgmt_port'], reply))
self.assertEqual(reply, '+OK', 'Failed to set slave_idle_timeout, REPLY=%s' % reply)
# Network isolation test
for cnt in range(5):
# Block network
util.log('\n\n\n ### BLOCK NETWORK, %d ### ' % cnt)
for s in cluster['servers']:
"""Loopback Address Range (Reference : RFC3330)
127.0.0.0/8 - This block is assigned for use as the Internet host
loopback address. A datagram sent by a higher level protocol to an
address anywhere within this block should loop back inside the host.
This is ordinarily implemented using only 127.0.0.1/32 for loopback,
but no addresses within this block should ever appear on any network
anywhere [RFC1700, page 5].
"""
self.assertTrue(util.iptables_drop('A', '127.0.0.100', s['smr_mgmt_port']), 'add a bloking role to iptables fail.')
for i in range(4):
util.log('waiting... %d' % (i + 1))
time.sleep(1)
# Check cluster state
ok = False
for i in range(7):
isolated_states = []
util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, isolated_states, check_quorum=True)
time.sleep(1)
state_transition_done = True
for s in isolated_states:
if s['ip'] != '127.0.0.100':
continue
if s['active_role'] != '?' or s['mgmt_role'] != 'N':
state_transition_done = False
if state_transition_done :
ok = True
break
time.sleep(1)
self.assertTrue(ok, 'Fail, state transition')
# Unblock network
util.log('\n\n\n ### UNBLOCK NETWORK, %d ### ' % cnt)
for s in cluster['servers']:
self.assertTrue(util.iptables_drop('D', '127.0.0.100', s['smr_mgmt_port']), 'delete a bloking role to iptables fail.')
# Check cluster state
ok = False
for i in range(7):
final_state = []
util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, final_state, check_quorum=True)
all_green = True
for s in final_state:
if is_pgs_normal(s) == False:
all_green = False
if all_green:
ok = True
break
time.sleep(1)
self.assertTrue(ok, 'Fail, state consistency')
# Check state
self.assertNotEqual(initial_state, None, 'initial_state is None')
self.assertNotEqual(final_state, None, 'final_state is None')
self.assertTrue(conf_checker.final_check())
# Shutdown cluster
default_cluster.finalize(cluster)
def test_6_repeat_isolation_and_no_opinion_linepay(self):
util.print_frame()
util.iptables_print_list()
# Add forwarding role
self.assertTrue(util.iptables_redirect('A', '127.0.0.100', '127.0.0.1'), 'add a forwarding role to iptables fail.')
self.assertTrue(util.iptables_redirect('A', '127.0.0.101', '127.0.0.1'), 'add a forwarding role to iptables fail.')
self.assertTrue(util.iptables_redirect('A', '127.0.0.102', '127.0.0.1'), 'add a forwarding role to iptables fail.')
cluster_name = 'no_opinion'
cluster = filter(lambda x: x['cluster_name'] == cluster_name, config.clusters)[0]
util.log(util.json_to_str(cluster))
self.leader_cm = cluster['servers'][0]
# MGMT
mgmt_ip = cluster['servers'][0]['real_ip']
mgmt_port = cluster['servers'][0]['cm_port']
# Create cluster
conf_checker = default_cluster.initialize_starting_up_smr_before_redis( cluster )
self.assertIsNotNone(conf_checker, 'failed to initialize cluster')
# Print initial state of cluster
util.log('\n\n\n ### INITIAL STATE OF CLUSTER ### ')
initial_state = []
self.assertTrue(util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, initial_state, check_quorum=True), 'failed to check cluster state')
# Network isolation test
loop_cnt = 0
while (loop_cnt < 20):
loop_cnt += 1
# Block network
util.log('\n\n\n ### BLOCK NETWORK, %d ### ' % loop_cnt)
self.assertTrue(util.iptables_drop('A', '127.0.0.102'), 'add a bloking role to iptables fail.')
for i in range(1):
util.log('waiting... %d' % (i + 1))
time.sleep(0.1)
self.assertTrue(util.iptables_drop('A', '127.0.0.100'), 'add a bloking role to iptables fail.')
for i in range(3):
util.log('waiting... %d' % (i + 1))
time.sleep(1.2)
self.assertTrue(util.iptables_drop('A', '127.0.0.101'), 'add a bloking role to iptables fail.')
for i in range(1):
util.log('waiting... %d' % (i + 1))
time.sleep(1)
# Unblock network
util.log('\n\n\n ### UNBLOCK NETWORK, %d ### ' % loop_cnt)
self.assertTrue(util.iptables_drop('D', '127.0.0.102'), 'delete a bloking role to iptables fail.')
for i in range(0):
util.log('waiting... %d' % (i + 1))
time.sleep(1)
self.assertTrue(util.iptables_drop('D', '127.0.0.100'), 'delete a bloking role to iptables fail.')
for i in range(0):
util.log('waiting... %d' % (i + 1))
time.sleep(1)
self.assertTrue(util.iptables_drop('D', '127.0.0.101'), 'delete a bloking role to iptables fail.')
for i in range(3):
util.log('waiting... %d' % (i + 1))
time.sleep(1)
# Print state of cluster
util.log('\n ### STATE OF CLUSTER ### ')
cluster_state = False
for i in range(10):
cluster_state = util.check_cluster(cluster_name, mgmt_ip, mgmt_port, initial_state, check_quorum=True)
if cluster_state == True:
break
else:
time.sleep(1)
self.assertTrue(cluster_state, 'failed to check cluster state')
all_in_f = True
for s in cluster['servers']:
if checkLastState(mgmt_ip, s['cm_port'], cluster_name, 0, 'F') == False:
all_in_f = False
break
self.assertFalse(all_in_f, 'PGS0`s last state remains in F')
# Delete forwarding role
self.assertTrue(util.iptables_redirect('D', '127.0.0.100', '127.0.0.1'), 'delete a forwarding role to iptables fail.')
self.assertTrue(util.iptables_redirect('D', '127.0.0.101', '127.0.0.1'), 'delete a forwarding role to iptables fail.')
self.assertTrue(util.iptables_redirect('D', '127.0.0.102', '127.0.0.1'), 'delete a forwarding role to iptables fail.')
self.assertTrue(conf_checker.final_check())
# Shutdown cluster
default_cluster.finalize(cluster)
def test_5_mgmt_is_isolated_with_lconn(self):
util.print_frame()
util.iptables_print_list()
cluster = filter(lambda x: x['cluster_name'] == 'network_isolation_cluster_1', config.clusters)[0]
util.log(util.json_to_str(cluster))
self.leader_cm = cluster['servers'][0]
# MGMT
mgmt_ip = cluster['servers'][0]['real_ip']
mgmt_port = cluster['servers'][0]['cm_port']
# Create cluster
conf_checker = default_cluster.initialize_starting_up_smr_before_redis( cluster )
self.assertIsNotNone(conf_checker, 'failed to initialize cluster')
# Print initial state of cluster
util.log('\n\n\n ### INITIAL STATE OF CLUSTER ### ')
initial_state = []
self.assertTrue(util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, initial_state, check_quorum=True), 'failed to check cluster state')
# Set SMR option (slave_idle_timeout)
util.log('\n\n\n ### Set SMR option ###')
for s in cluster['servers']:
t = telnet.Telnet('SMR%d' % s['id'])
self.assertEqual(t.connect(s['ip'], s['smr_mgmt_port']), 0,
'Failed to connect to smr. ADDR=%s:%d' % (s['ip'], s['smr_mgmt_port']))
cmd = 'confset slave_idle_timeout_msec 18000'
util.log('[%s:%d] >> %s' % (s['ip'], s['smr_mgmt_port'], cmd))
t.write('confset slave_idle_timeout_msec 18000\r\n')
reply = t.read_until('\r\n').strip()
util.log('[%s:%d] << %s' % (s['ip'], s['smr_mgmt_port'], reply))
self.assertEqual(reply, '+OK', 'Failed to set slave_idle_timeout, REPLY=%s' % reply)
# Network isolation test
for loop_cnt in range(3):
# Get master
master = util.get_server_by_role_and_pg( cluster['servers'], 'master', 0 )
first_slave = None
for s in cluster['servers']:
if s == master:
continue
# Skip non-virtual host
if s.has_key('real_ip') == False:
continue
if first_slave == None:
first_slave = s
# 'role lconn'
util.log( 'role lconn pgs%d while hanging.' % s['id'] )
ret = util.role_lconn_addr( s['real_ip'], s['smr_mgmt_port'] )
self.assertEqual( ret, '+OK\r\n', 'role lconn failed. reply="%s"' % (ret[:-2]) )
util.log( 'succeeded : cmd="role lconn", reply="%s"' % (ret[:-2]) )
time.sleep(0.5)
# Block network
util.log('\n\n\n ### BLOCK NETWORK, %d ### ' % loop_cnt)
self.assertTrue(util.iptables_drop('A', '127.0.0.100', first_slave['smr_mgmt_port']), 'add a bloking role to iptables fail.')
for i in range(6):
util.log('waiting... %d' % (i + 1))
time.sleep(1)
# Check cluster state
ok = False
for i in range(10):
isolated_states = []
util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, isolated_states, check_quorum=True)
time.sleep(1)
state_transition_done = True
for s in isolated_states:
if s['pgs_id'] != first_slave['id']:
continue
if s['active_role'] != '?' or s['mgmt_role'] != 'N':
state_transition_done = False
if state_transition_done :
ok = True
break
time.sleep(1)
self.assertTrue(ok, 'Fail, state transition')
# Unblock network
self.assertTrue(util.iptables_drop('D', '127.0.0.100', first_slave['smr_mgmt_port']), 'delete a bloking role to iptables fail.')
# Check cluster state
ok = False
for i in range(7):
final_state = []
util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, final_state, check_quorum=True)
state_consistency = True
for s in final_state:
if s['pgs_id'] == 1:
continue
if is_pgs_normal(s) == False:
state_consistency = False
if state_consistency:
ok = True
break
time.sleep(1)
self.assertTrue(ok, 'Fail, state consistency')
ok = False
for i in xrange(5):
ok = util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, check_quorum=True)
if ok:
break
else:
time.sleep(1)
self.assertTrue(ok, 'failed to check cluster state')
# Check state
self.assertNotEqual(initial_state, None, 'initial_state is None')
self.assertNotEqual(final_state, None, 'final_state is None')
initial_state = sorted(initial_state, key=lambda x: int(x['pgs_id']))
final_state = sorted(final_state, key=lambda x: int(x['pgs_id']))
for i in range(len(final_state)):
msg = 'ts (%d)%d -> (%d)%d' % (initial_state[i]['pgs_id'], initial_state[i]['active_ts'], final_state[i]['pgs_id'], final_state[i]['active_ts'])
util.log(msg)
self.assertTrue(util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, check_quorum=True), 'failed to check cluster state')
self.assertTrue(conf_checker.final_check())
# Shutdown cluster
default_cluster.finalize(cluster)
def test_4_mgmt_is_isolated_with_red_failover(self):
util.print_frame()
util.iptables_print_list()
cluster = filter(lambda x: x['cluster_name'] == 'network_isolation_cluster_1', config.clusters)[0]
util.log(util.json_to_str(cluster))
self.leader_cm = cluster['servers'][0]
# MGMT
mgmt_ip = cluster['servers'][0]['real_ip']
mgmt_port = cluster['servers'][0]['cm_port']
# Create cluster
conf_checker = default_cluster.initialize_starting_up_smr_before_redis( cluster )
self.assertIsNotNone(conf_checker, 'failed to initialize cluster')
util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port)
# Master must be the first pgs, cluster['servers'][0].
to_be_master = cluster['servers'][0]
m = util.get_server_by_role_and_pg(cluster['servers'], 'master', to_be_master['pg_id'])
master_id = -1
if m['id'] != to_be_master['id']:
try_cnt = 0
while master_id != to_be_master['id'] and try_cnt < 20:
master_id = util.role_change(cluster['servers'][0], cluster['cluster_name'], to_be_master['id'])
try_cnt += 1
time.sleep(1)
self.assertEquals(master_id, to_be_master['id'], 'change %d to a master fail' % to_be_master['id'])
# Print initial state of cluster
util.log('\n\n\n ### INITIAL STATE OF CLUSTER ### ')
initial_state = []
self.assertTrue(util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, initial_state, check_quorum=True), 'failed to check cluster state')
# Set SMR option (slave_idle_timeout)
util.log('\n\n\n ### Set SMR option ###')
for s in cluster['servers']:
t = telnet.Telnet('SMR%d' % s['id'])
self.assertEqual(t.connect(s['ip'], s['smr_mgmt_port']), 0,
'Failed to connect to smr. ADDR=%s:%d' % (s['ip'], s['smr_mgmt_port']))
cmd = 'confset slave_idle_timeout_msec 18000'
util.log('[%s:%d] >> %s' % (s['ip'], s['smr_mgmt_port'], cmd))
t.write('confset slave_idle_timeout_msec 18000\r\n')
reply = t.read_until('\r\n').strip()
util.log('[%s:%d] << %s' % (s['ip'], s['smr_mgmt_port'], reply))
self.assertEqual(reply, '+OK', 'Failed to set slave_idle_timeout, REPLY=%s' % reply)
# Network isolation test
for loop_cnt in range(3):
# Block network
util.log('\n\n\n ### BLOCK NETWORK, %d ### ' % loop_cnt)
for s in cluster['servers']:
self.assertTrue(util.iptables_drop('A', '127.0.0.100', s['smr_mgmt_port']), 'add a bloking role to iptables fail.')
for i in range(4):
util.log('waiting... %d' % (i + 1))
time.sleep(1)
# Check cluster state
ok = False
for i in range(7):
isolated_states = []
util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, isolated_states, check_quorum=True)
time.sleep(1)
state_transition_done = True
for s in isolated_states:
if s['ip'] != '127.0.0.100':
continue
if s['active_role'] != '?' or s['mgmt_role'] != 'N':
state_transition_done = False
if state_transition_done :
ok = True
break
time.sleep(1)
self.assertTrue(ok, 'Fail, state transition')
pgs_list = util.get_pgs_info_list(mgmt_ip, mgmt_port, cluster)
reds = filter(lambda x: x['color'] == 'RED', pgs_list)
# Shutdown
server = cluster['servers'][random.choice(reds)['pgs_id']]
util.log( 'shutdown pgs%d while hanging.' % server['id'] )
ret = testbase.request_to_shutdown_smr( server )
self.assertEqual( ret, 0, 'failed to shutdown smr. id:%d' % server['id'] )
ret = testbase.request_to_shutdown_redis( server )
self.assertEqual( ret, 0, 'failed to shutdown redis. id:%d' % server['id'] )
# Check state F
max_try = 20
expected = 'F'
for i in range( 0, max_try):
util.log('MGMT_IP:%s, MGMT_PORT:%d' % (mgmt_ip, mgmt_port))
state = util._get_smr_state( server['id'], cluster['cluster_name'], mgmt_ip, mgmt_port )
if expected == state:
break;
time.sleep( 1 )
self.assertEqual( expected , state,
'server%d - state:%s, expected:%s' % (server['id'], state, expected) )
util.log( 'succeeded : pgs%d state changed to F.' % server['id'] )
# Unblock network
for s in cluster['servers']:
self.assertTrue(util.iptables_drop('D', '127.0.0.100', s['smr_mgmt_port']), 'delete a bloking role to iptables fail.')
# Check cluster state
ok = False
for i in range(10):
final_state = []
util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, final_state, check_quorum=True)
state_consistency = True
for s in final_state:
if s['pgs_id'] == server['id']:
continue
if is_pgs_normal(s) == False:
state_consistency = False
if state_consistency:
ok = True
break
time.sleep(1)
self.assertTrue(ok, 'Fail, state consistency')
# Recovery
util.log( 'restart pgs%d.' % server['id'] )
ret = testbase.request_to_start_smr( server )
self.assertEqual( ret, 0, 'failed to start smr. id:%d' % server['id'] )
ret = testbase.request_to_start_redis( server )
self.assertEqual( ret, 0, 'failed to start redis. id:%d' % server['id'] )
wait_count = 20
ret = testbase.wait_until_finished_to_set_up_role( server, wait_count )
self.assertEqual( ret, 0, 'failed to role change. smr_id:%d' % (server['id']) )
redis = redis_mgmt.Redis( server['id'] )
ret = redis.connect( server['ip'], server['redis_port'] )
self.assertEqual( ret, 0, 'failed to connect to redis' )
ok = False
for i in xrange(5):
ok = util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, check_quorum=True)
if ok:
break
else:
time.sleep(1)
self.assertTrue(ok, 'failed to check cluster state')
# Reset SMR option (slave_idle_timeout)
t = telnet.Telnet('SMR%d' % server['id'])
self.assertEqual(t.connect(server['ip'], server['smr_mgmt_port']), 0,
'Failed to connect to smr. ADDR=%s:%d' % (server['ip'], server['smr_mgmt_port']))
cmd = 'confset slave_idle_timeout_msec 18000'
util.log('[%s:%d] >> %s' % (server['ip'], server['smr_mgmt_port'], cmd))
t.write('confset slave_idle_timeout_msec 18000\r\n')
reply = t.read_until('\r\n').strip()
util.log('[%s:%d] << %s' % (server['ip'], server['smr_mgmt_port'], reply))
self.assertEqual(reply, '+OK', 'Failed to set slave_idle_timeout, REPLY=%s' % reply)
# Check state
self.assertNotEqual(initial_state, None, 'initial_state is None')
self.assertNotEqual(final_state, None, 'final_state is None')
initial_state = sorted(initial_state, key=lambda x: int(x['pgs_id']))
final_state = sorted(final_state, key=lambda x: int(x['pgs_id']))
for i in range(len(final_state)):
msg = 'ts (%d)%d -> (%d)%d' % (initial_state[i]['pgs_id'], initial_state[i]['active_ts'], final_state[i]['pgs_id'], final_state[i]['active_ts'])
util.log(msg)
if initial_state[i]['pgs_id'] == 1:
self.assertNotEqual(initial_state[i]['active_ts'], final_state[i]['active_ts'], msg)
self.assertTrue(util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, check_quorum=True), 'failed to check cluster state')
self.assertTrue(conf_checker.final_check())
# Shutdown cluster
default_cluster.finalize(cluster)
def test_3_some_pgs_is_isolated_2copy(self):
util.print_frame()
out = util.sudo('iptables -L')
util.log('====================================================================')
util.log('out : %s' % out)
util.log('out.return_code : %d' % out.return_code)
util.log('out.stderr : %s' % out.stderr)
util.log('out.succeeded : %s' % out.succeeded)
# Add forwarding role (127.0.0.100 -> 127.0.0.1)
self.assertTrue(util.iptables_redirect('A', '127.0.0.100', '127.0.0.1'), 'add a forwarding role to iptables fail.')
cluster = filter(lambda x: x['cluster_name'] == 'network_isolation_cluster_1_2copy', config.clusters)[0]
util.log(util.json_to_str(cluster))
# MGMT
mgmt_ip = cluster['servers'][0]['ip']
mgmt_port = cluster['servers'][0]['cm_port']
# Create cluster
conf_checker = default_cluster.initialize_starting_up_smr_before_redis( cluster )
self.assertIsNotNone(conf_checker, 'failed to initialize cluster')
# Place master on real ip address
for pg_id in [0, 1]:
m = util.get_server_by_role_and_pg(cluster['servers'], 'master', pg_id)
s = util.get_server_by_role_and_pg(cluster['servers'], 'slave', pg_id)
if m.has_key('ip') and m.has_key('real_ip'):
if m['ip'] != m['real_ip']:
ret = util.role_change(cluster['servers'][0], cluster['cluster_name'], s['id'])
self.assertNotEquals(ret, -1, 'change %d to a master fail' % s['id'])
# Print initial state of cluster
util.log('\n\n\n ### INITIAL STATE OF CLUSTER ### ')
initial_state = []
self.assertTrue(util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, initial_state, check_quorum=True), 'failed to check cluster state')
# Network isolation test
for cnt in range(3):
# Block network
util.log('\n\n\n ### BLOCK NETWORK, %d ### ' % cnt)
self.assertTrue(util.iptables_drop('A', '127.0.0.100'), 'add a bloking role to iptables fail.')
for i in range(4):
util.log('waiting... %d' % (i + 1))
time.sleep(1)
# Check cluster state
ok = False
for i in range(7):
isolated_states = []
util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, isolated_states, check_quorum=True)
time.sleep(1)
state_transition_done = True
for s in isolated_states:
if s['pgs_id'] == 0 or s['pgs_id'] == 1:
continue
if s['active_role'] != 'M' or s['mgmt_role'] != 'M':
state_transition_done = False
if s['quorum'] != 0:
state_transition_done = False
if state_transition_done:
ok = True
break
time.sleep(1)
self.assertTrue(ok, 'Fail, state transition')
# Unblock network
util.log('\n\n\n ### UNBLOCK NETWORK, %d ### ' % cnt)
self.assertTrue(util.iptables_drop('D', '127.0.0.100'), 'delete a bloking role to iptables fail.')
# Check cluster state
ok = False
for i in range(7):
final_state = []
if util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, final_state, check_quorum=True) == False:
time.sleep(1)
continue
state_consistency = True
for s in final_state:
if s['pgs_id'] == 1:
continue
if is_pgs_normal(s) == False:
state_consistency = False
if state_consistency:
ok = True
break
time.sleep(1)
self.assertTrue(ok, 'Fail, state consistency')
# Check state
self.assertNotEqual(initial_state, None, 'initial_state is None')
self.assertNotEqual(final_state, None, 'final_state is None')
# Delete forwarding role (127.0.0.100 -> 127.0.0.1)
self.assertTrue(util.iptables_redirect('D', '127.0.0.100', '127.0.0.1'), 'delete a forwarding role to iptables fail.')
self.assertTrue(conf_checker.final_check())
# Shutdown cluster
default_cluster.finalize(cluster)
def test_2_some_pgs_is_isolated(self):
util.print_frame()
util.iptables_print_list()
# Add forwarding role (127.0.0.100 -> 127.0.0.1)
self.assertTrue(util.iptables_redirect('A', '127.0.0.100', '127.0.0.1'), 'add a forwarding role to iptables fail.')
cluster = filter(lambda x: x['cluster_name'] == 'network_isolation_cluster_2', config.clusters)[0]
util.log(util.json_to_str(cluster))
# MGMT
mgmt_ip = cluster['servers'][0]['real_ip']
mgmt_port = cluster['servers'][0]['cm_port']
# Create cluster
conf_checker = default_cluster.initialize_starting_up_smr_before_redis( cluster )
self.assertIsNotNone(conf_checker, 'failed to initialize cluster')
# Place master on virtual ip address in order to cause master election.
pg_id = 0
m = util.get_server_by_role_and_pg(cluster['servers'], 'master', pg_id)
s = util.get_server_by_role_and_pg(cluster['servers'], 'slave', pg_id)
if m.has_key('ip') == True and m.has_key('real_ip') == False:
ret = util.role_change(cluster['servers'][0], cluster['cluster_name'], s['id'])
self.assertNotEquals(ret, -1, 'change %d to a master fail' % s['id'])
# Print initial state of cluster
util.log('\n\n\n ### INITIAL STATE OF CLUSTER ### ')
initial_state = []
self.assertTrue(util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, initial_state, check_quorum=True), 'failed to check cluster state')
# Network isolation test
for cnt in range(3):
# Block network
util.log('\n\n\n ### BLOCK NETWORK, %d ### ' % cnt)
self.assertTrue(util.iptables_drop('A', '127.0.0.100'), 'add a bloking role to iptables fail.')
for i in range(4):
util.log('waiting... %d' % (i + 1))
time.sleep(1)
# Check cluster state
ok = False
for i in range(7):
isolated_states = []
util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, isolated_states, check_quorum=True)
time.sleep(1)
state_transition_done = True
for s in isolated_states:
if s['pgs_id'] == 1:
continue
if s['active_role'] != '?' or s['mgmt_role'] != 'N':
state_transition_done = False
if state_transition_done :
ok = True
break
time.sleep(1)
self.assertTrue(ok, 'Fail, state transition')
# Unblock network
util.log('\n\n\n ### UNBLOCK NETWORK, %d ### ' % cnt)
self.assertTrue(util.iptables_drop('D', '127.0.0.100'), 'delete a bloking role to iptables fail.')
# Check cluster state
ok = False
for i in range(7):
final_state = []
util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, final_state, check_quorum=True)
state_consistency = True
for s in final_state:
if s['pgs_id'] == 1:
continue
if is_pgs_normal(s) == False:
state_consistency = False
if state_consistency:
ok = True
break
time.sleep(1)
self.assertTrue(ok, 'Fail, state consistency')
# Check state
self.assertNotEqual(initial_state, None, 'initial_state is None')
self.assertNotEqual(final_state, None, 'final_state is None')
initial_state = sorted(initial_state, key=lambda x: int(x['pgs_id']))
final_state = sorted(final_state, key=lambda x: int(x['pgs_id']))
for i in range(len(final_state)):
msg = 'ts (%d)%d -> (%d)%d' % (initial_state[i]['pgs_id'], initial_state[i]['active_ts'], final_state[i]['pgs_id'], final_state[i]['active_ts'])
util.log(msg)
self.assertNotEqual(initial_state[i]['active_ts'], final_state[i]['active_ts'], msg)
# Delete forwarding role (127.0.0.100 -> 127.0.0.1)
self.assertTrue(util.iptables_redirect('D', '127.0.0.100', '127.0.0.1'), 'delete a forwarding role to iptables fail')
self.assertTrue(conf_checker.final_check())
# Shutdown cluster
default_cluster.finalize(cluster)
def test_gateway_network_isolation(self):
util.print_frame()
cluster = self.cluster
# Clear rules
while True:
if util.iptables_redirect('D', '127.0.0.100', '127.0.0.1') == False:
break
while True:
if util.iptables_drop('D', '127.0.0.100') == False:
break
# Print rules
util.iptables_print_list()
# Start loadgenerators
self.load_gen_list = {}
for i in range(len(cluster['servers'])):
arc_api = ARC_API(ZK_ADDR, CLUSTER_NAME, logFilePrefix = self.arcci_log, so_path = self.so_path)
server = cluster['servers'][i]
load_gen = LoadGenerator_ARCCI_FaultTolerance(server['id'], arc_api)
load_gen.start()
self.load_gen_list[i] = load_gen
# Add forwarding role (127.0.0.100 -> 127.0.0.1)
self.assertTrue(util.iptables_redirect('A', '127.0.0.100', '127.0.0.1'), 'add a forwarding role to iptables fail.')
# Add virtualhost information to MGMT
VIRTUAL_HOST_NAME = 'virtualhost'
VIRTUAL_HOST_IP = '127.0.0.100'
ret = util.pm_add(VIRTUAL_HOST_NAME, VIRTUAL_HOST_IP, MGMT_IP, MGMT_PORT)
self.assertTrue(ret, 'pm_add fail.')
# Modify gateway information of MGMT
server = cluster['servers'][0]
gw_id = server['id']
gw_port = server['gateway_port']
# Delete gateway
ret = util.gw_del(CLUSTER_NAME, gw_id, MGMT_IP, MGMT_PORT)
self.assertTrue(ret, 'gw_del fail')
# Add gateway
ret= util.gw_add(CLUSTER_NAME, gw_id, VIRTUAL_HOST_NAME, VIRTUAL_HOST_IP, gw_port, MGMT_IP, MGMT_PORT)
self.assertTrue(ret, 'gw_add fail')
# Check load balancing
for i in range(5):
ok = True
for s in cluster['servers']:
tps = util.get_tps(s['ip'], s['gateway_port'], 'gw')
util.log('%s:%d TPS:%d' % (s['ip'], s['gateway_port'], tps))
if tps < 50:
ok = False
if ok:
break
time.sleep(1)
self.assertTrue(ok, 'load balancing fail')
util.log('load balancing success')
# Block
self.assertTrue(util.iptables_drop('A', '127.0.0.100'), 'add a bloking role to iptables fail.')
# Check blocked gateway`s ops
for i in range(5):
ok = True
tps = util.get_tps(server['ip'], server['gateway_port'], 'gw')
util.log('%s:%d TPS:%d' % (server['ip'], server['gateway_port'], tps))
if tps > 10:
ok = False
if ok:
break
time.sleep(1)
self.assertTrue(ok, 'load balancing fail - blocked gateway')
util.log('load balancing success - blocked gateway')
# Check unblocked gateway`s ops
for i in range(10):
ok = True
for s in cluster['servers']:
if s == server:
continue
tps = util.get_tps(s['ip'], s['gateway_port'], 'gw')
util.log('%s:%d TPS:%d' % (s['ip'], s['gateway_port'], tps))
if tps < 50:
ok = False
if ok:
break
time.sleep(1)
self.assertTrue(ok, 'load balancing fail - nonblocked gateways')
util.log('load balancing success - nonblocked gateways')
# Unblock
self.assertTrue(util.iptables_drop('D', '127.0.0.100'), 'delete a bloking role to iptables fail.')
# Check load balancing
ok = False
for i in xrange(5):
condition = (lambda s: (s['ops'] <= 10 if s['id'] == gw_id else s['ops'] >= 50))
if util.check_ops(cluster['servers'], 'gw', condition):
ok = True
break
time.sleep(1)
self.assertTrue(ok, 'load balancing fail - all gateways after unblocking network')
util.log('load balancing success - all gateways after unblocking network')
server = cluster['servers'][0]
# Wait until opinion for the gateway deleted.
for i in xrange(5):
util.log('Wait until opinions for the gateway have been deleted... %d' % i)
time.sleep(1)
# Delete gateway
ret = util.gw_del(CLUSTER_NAME, gw_id, MGMT_IP, MGMT_PORT)
self.assertTrue(ret, 'gw_del fail')
# Add gateway
ret= util.gw_add(CLUSTER_NAME, gw_id, server['pm_name'], server['ip'], gw_port, MGMT_IP, MGMT_PORT)
self.assertTrue(ret, 'gw_add fail')
# Check load balancing
for i in range(10):
ok = True
for s in cluster['servers']:
tps = util.get_tps(s['ip'], s['gateway_port'], 'gw')
util.log('%s:%d TPS:%d' % (s['ip'], s['gateway_port'], tps))
if tps < 50:
ok = False
if ok:
break
time.sleep(1)
self.assertTrue(ok, 'load balancing fail - all gateways after unblocking network')
util.log('load balancing success - all gateways after unblocking network')
# Go back to initial configuration
self.assertTrue(util.pm_del(MGMT_IP, MGMT_PORT, VIRTUAL_HOST_NAME),
'failed to pm_del to go back to initial configuration.')