def process(self, args):
action = args.get('action')
result = brick.execute(self.config, action, args)
ret = util.json_to_str(result)
print(ret[:1000])
return ret
def _gateway_load_info(self, api):
time.sleep(3)
log_reader = LogReader(api.conf.log_file_prefix)
gw_list = copy.deepcopy(self.GW_LIST)
while True:
line = log_reader.readline()
if line == None:
break
if line.find(MSG_GATEWAY_ADD_ZK) != -1:
gw = line.split('data:')[1]
gw = ast.literal_eval(gw)
if gw not in gw_list:
self.fail("FAIL, unexpected gateway information. gw:%s" % util.json_to_str(gw))
else:
print gw
gw_list.remove(gw)
continue
else:
for gw in gw_list:
find_str = MSG_GATEWAY_ADD_GW_PREFIX_FMT % (gw['ip'], gw['port'])
if line.find(find_str) != -1:
if line.find(MSG_GATEWAY_ADD_GW_POSTFIX) != -1:
print gw
gw_list.remove(gw)
if len(gw_list) != 0:
self.fail("FAIL, load gateway information. gw:%s" % util.json_to_str(gw_list))
else:
util.log('SUCCESS, load gateway information.')
api.destroy()
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_delete_smrlog_after_scaleout(self):
util.print_frame()
# start load generator
util.log("start load_generator")
for i in range(self.max_load_generator):
ip, port = util.get_rand_gateway(self.cluster)
self.load_gen_thrd_list[i] = load_generator.LoadGenerator(i, ip, port)
self.load_gen_thrd_list[i].start()
time.sleep(5) # generate load for 5 sec
util.log("started load_generator")
# servers for scale out
servers = [config.server4, config.server5, config.server6]
leader_cm = self.cluster['servers'][0]
# Scale out
cluster = config.clusters[0]
ret = util.pg_add(cluster, servers, leader_cm)
self.assertEqual(True, ret, 'Scale out fail. util.pg_add returns false')
time.sleep(5)
# pg0 -> pg1
cluster = config.clusters[1]
ret = util.migration(cluster, 0, 1, 8000, 8191, 40000)
self.assertEqual(True, ret, 'Migration Fail 0 -> 1')
# get log file
old_logs = {}
for s in config.clusters[0]['servers']:
parent_dir, log_dir = util.smr_log_dir(s['id'])
path = '%s/%s' % (parent_dir, log_dir)
old_logs[s['id']] = util.ls(path)
# bgsave in order to make smrlogs deleted.
for s in config.clusters[0]['servers']:
bgsave_ret = util.bgsave(s)
self.assertTrue(bgsave_ret, 'failed to bgsave. pgs%d' % s['id'])
util.log('bgsave pgs%d is done.')
# check consistency
ok = True
for j in range(len(self.load_gen_thrd_list)):
self.assertTrue(self.load_gen_thrd_list[j].isConsistent(),
'Inconsistent after migration')
# is smr-replicator delete smrlogs?
i = 0
while i < 20:
i += 1
# get current log files
cur_logs = {}
for s in config.clusters[0]['servers']:
parent_dir, log_dir = util.smr_log_dir(s['id'])
path = '%s/%s' % (parent_dir, log_dir)
cur_logs[s['id']] = util.ls(path)
# compare old and new
temp_old_logs = copy.deepcopy(old_logs)
for id, nl in cur_logs.items():
ol = temp_old_logs.get(id)
self.assertNotEqual(ol, None, "failed to check logfiles. old logs for smr-replicator '%d' is not exist." % id)
for log in nl:
if log in ol:
ol.remove(log)
ok = True
for id, ol in temp_old_logs.items():
if len(ol) == 0:
ok = False
util.log('Loop %d ---------------------------------------------------------' % i)
util.log('deleted smrlog files: %s' % util.json_to_str(temp_old_logs))
if ok:
break
time.sleep(10)
self.assertTrue(ok, 'smr-replicator does not delete smrlogs.')
util.log('smr-replicator deletes smrlogs.')
# check consistency of load_generator
for i in range(len(self.load_gen_thrd_list)):
self.load_gen_thrd_list[i].quit()
for i in range(len(self.load_gen_thrd_list)):
self.load_gen_thrd_list[i].join()
self.assertTrue(self.load_gen_thrd_list[i].isConsistent(), 'Inconsistent after migration')
def test_delete_smrlog_after_scaleout(self):
util.print_frame()
# start load generator
util.log("start load_generator")
for i in range(self.max_load_generator):
ip, port = util.get_rand_gateway(self.cluster)
self.load_gen_thrd_list[i] = load_generator.LoadGenerator(
i, ip, port)
self.load_gen_thrd_list[i].start()
time.sleep(5) # generate load for 5 sec
util.log("started load_generator")
# servers for scale out
servers = [config.server4, config.server5, config.server6]
leader_cm = self.cluster['servers'][0]
# Scale out
cluster = config.clusters[0]
ret = util.pg_add(cluster, servers, leader_cm)
self.assertEqual(True, ret,
'Scale out fail. util.pg_add returns false')
time.sleep(5)
# pg0 -> pg1
cluster = config.clusters[1]
ret = util.migration(cluster, 0, 1, 8000, 8191, 40000)
self.assertEqual(True, ret, 'Migration Fail 0 -> 1')
# get log file
old_logs = {}
for s in config.clusters[0]['servers']:
parent_dir, log_dir = util.smr_log_dir(s['id'])
path = '%s/%s' % (parent_dir, log_dir)
old_logs[s['id']] = util.ls(path)
# bgsave in order to make smrlogs deleted.
for s in config.clusters[0]['servers']:
bgsave_ret = util.bgsave(s)
self.assertTrue(bgsave_ret, 'failed to bgsave. pgs%d' % s['id'])
util.log('bgsave pgs%d is done.')
# check consistency
ok = True
for j in range(len(self.load_gen_thrd_list)):
self.assertTrue(self.load_gen_thrd_list[j].isConsistent(),
'Inconsistent after migration')
# is smr-replicator delete smrlogs?
i = 0
while i < 20:
i += 1
# get current log files
cur_logs = {}
for s in config.clusters[0]['servers']:
parent_dir, log_dir = util.smr_log_dir(s['id'])
path = '%s/%s' % (parent_dir, log_dir)
cur_logs[s['id']] = util.ls(path)
# compare old and new
temp_old_logs = copy.deepcopy(old_logs)
for id, nl in cur_logs.items():
ol = temp_old_logs.get(id)
self.assertNotEqual(
ol, None,
"failed to check logfiles. old logs for smr-replicator '%d' is not exist."
% id)
for log in nl:
if log in ol:
ol.remove(log)
ok = True
for id, ol in temp_old_logs.items():
if len(ol) == 0:
ok = False
util.log(
'Loop %d ---------------------------------------------------------'
% i)
util.log('deleted smrlog files: %s' %
util.json_to_str(temp_old_logs))
if ok:
break
time.sleep(10)
self.assertTrue(ok, 'smr-replicator does not delete smrlogs.')
util.log('smr-replicator deletes smrlogs.')
# check consistency of load_generator
for i in range(len(self.load_gen_thrd_list)):
self.load_gen_thrd_list[i].quit()
for i in range(len(self.load_gen_thrd_list)):
self.load_gen_thrd_list[i].join()
self.assertTrue(self.load_gen_thrd_list[i].isConsistent(),
'Inconsistent after migration')
def test_2_some_pgs_is_isolated(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)
out = util.sudo('iptables -t nat -A OUTPUT -d 127.0.0.100 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'add a forwarding role to iptables fail. output:%s' % out)
out = util.sudo('iptables -t nat -A PREROUTING -d 127.0.0.100 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'add a forwarding role to iptables fail. output:%s' % out)
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
ret = default_cluster.initialize_starting_up_smr_before_redis( cluster )
self.assertEqual(0, ret, 'failed to TestMaintenance.initialize')
# 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)
out = util.sudo('iptables -A OUTPUT -d 127.0.0.100 -j DROP')
self.assertTrue(out.succeeded, 'add a bloking role to iptables fail. output:%s' % out)
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)
out = util.sudo('iptables -D OUTPUT -d 127.0.0.100 -j DROP')
self.assertTrue(out.succeeded, 'delete a bloking role to iptables fail. output:%s' % out)
# 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 s['active_role'] != s['mgmt_role']:
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)
# Shutdown cluster
ret = default_cluster.finalize( cluster )
self.assertEqual(ret, 0, 'failed to TestMaintenance.finalize')
# Delete forwarding role (127.0.0.100 -> 127.0.0.1)
out = util.sudo('iptables -t nat -D OUTPUT -d 127.0.0.100 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'delete a forwarding role to iptables fail. output:%s' % out)
out = util.sudo('iptables -t nat -D PREROUTING -d 127.0.0.100 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'delete a forwarding role to iptables fail. output:%s' % out)
def test_7_dirty_network_fi(self):
util.print_frame()
clnts = []
try:
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.')
cluster_name = 'network_isolation_cluster_1'
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,
conf={'cm_context':'applicationContext-fi.xml'})
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')
# Start crc16 client
for s in cluster['servers']:
c = load_generator_crc16.Crc16Client(s['id'], s['ip'], s['redis_port'], 600, verbose=False)
c.start()
clnts.append(c)
# Network isolation test
cmfi = fi_confmaster.ConfmasterWfFi(['ra', 'me', 'yj', 'bj', 'mg'],
['lconn', 'slave', 'master', 'setquorum'], [True, False], 1)
for fi in cmfi:
# Block network
util.log('\n\n\n ### BLOCK NETWORK, %s ### ' % str(fi))
ret = block_network(cluster, mgmt_ip, mgmt_port)
self.assertTrue(ret, '[%s] failed to block network.' % str(fi))
for i in xrange(4):
util.log('waiting... %d' % (i + 1))
time.sleep(1)
# Check cluster state
ok = False
for i in xrange(10):
isolated_states = []
util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, isolated_states, check_quorum=True)
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')
# Fault injection
try:
self.assertTrue(fi_confmaster.fi_add(fi, 1, mgmt_ip, mgmt_port),
"Confmaster command fail. fi: %s" % str(fi))
except ValueError as e:
self.fail("Confmaster command error. cmd: \"%s\", reply: \"%s\"" % (cmd, reply))
# Unblock network
util.log('\n\n\n ### UNBLOCK NETWORK, %s ### ' % str(fi))
ret = unblock_network(cluster, mgmt_ip, mgmt_port, None)
self.assertTrue(ret, '[%s] failed to unblock network.' % str(fi))
for i in xrange(4):
util.log('waiting... %d' % (i + 1))
time.sleep(1)
# Check cluster state
ok = False
for i in xrange(10):
isolated_states = []
ok = util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, isolated_states, check_quorum=True)
if ok:
break
time.sleep(1)
self.assertTrue(ok, '[%s] Fail. unstable cluster.' % str(fi))
check_cluster = False
# 'bj', 'slave'
if fi[0] == 'bj' and fi[1] == 'slave':
m, s1, s2 = util.get_mss(cluster)
ret = util.role_lconn(s1)
self.assertEqual("+OK\r\n", ret, '[%s] role lconn fail.' % str(fi))
check_cluster = True
# 'me', 'lconn'
elif fi[0] == 'me':
m, s1, s2 = util.get_mss(cluster)
ret = util.role_lconn(m)
self.assertEqual("+OK\r\n", ret, '[%s] role lconn fail.' % str(fi))
check_cluster = True
# 'setquorum'
elif fi[1] == 'setquorum':
m, s1, s2 = util.get_mss(cluster)
ret = util.cmd_to_smr_addr(s1['ip'], s1['smr_mgmt_port'], 'fi delay sleep 1 8000\r\n', timeout=20)
self.assertEqual("+OK\r\n", ret, '[%s] "fi delay sleep 1 8000" fail. ret: "%s"' % (str(fi), ret))
check_cluster = True
if check_cluster:
# Check cluster state
ok = False
for i in xrange(20):
isolated_states = []
ok = util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, isolated_states, check_quorum=True)
if ok:
break
time.sleep(1)
self.assertTrue(ok, '[%s] Fail. unstable cluster.' % str(fi))
# Check fault injection
ok = False
for i in xrange(10):
count = fi_confmaster.fi_count(fi, mgmt_ip, mgmt_port)
if count == 0:
ok = True
break
time.sleep(0.5)
self.assertTrue(ok, "[%s] fail. failt injection had not been triggered." % str(fi))
for c in clnts:
self.assertTrue(c.is_consistency(), '[%s] data consistency error!' % str(fi))
for c in clnts:
self.assertTrue(c.is_consistency(), '[%s] data consistency error!' % str(fi))
# Go back to initial configuration
cmfi.init()
for fi in cmfi:
try:
self.assertTrue(fi_confmaster.fi_add(fi, 0, mgmt_ip, mgmt_port),
"Confmaster command fail. fi: %s" % str(fi))
except ValueError as e:
self.fail("Confmaster command error. cmd: \"%s\", reply: \"%s\"" % (cmd, reply))
# Wait until workflows done
ret = util.await(60, True)(
lambda cinfo: cinfo['wf'] == 0,
lambda : util.cluster_info(mgmt_ip, mgmt_port, cluster['cluster_name']))
self.assertTrue(ret, 'There are still some workflows.')
self.assertTrue(conf_checker.final_check())
# Shutdown cluster
default_cluster.finalize(cluster)
finally:
for c in clnts:
c.quit()
for c in clnts:
c.join()
# Delete forwarding role
self.assertTrue(util.iptables_redirect('D', '127.0.0.100', '127.0.0.1'), 'add a forwarding role to iptables fail.')
def test_gateway_add_del(self):
util.print_frame()
api = ARC_API(ZK_ADDR, CLUSTER_NAME, logFilePrefix = self.arcci_log, so_path = self.so_path)
# Add gateway
gw_port = 10000
gw_id = 10
cmd = 'gw_add %s %d %s %s %d' % (CLUSTER_NAME, gw_id, HOST_NAME, HOST_IP, gw_port)
ret = util.cm_command(MGMT_IP, MGMT_PORT, cmd)
if ret != None and len(ret) >= 2:
ret = ret[:-2]
util.log('cmd:"%s", ret:"%s"' % (cmd, ret))
if not ret.startswith('{"state":"success","msg":"+OK"}'):
self.fail('failed to add gateway')
# Deploy gateway
server = self.cluster['servers'][0]
ret = util.deploy_gateway(gw_id)
self.assertTrue(ret, 'failed to deploy_gateway')
# Start gateway
ret = util.start_gateway( gw_id, server['ip'], MGMT_PORT, server['cluster_name'], gw_port)
self.assertEqual( ret, 0, 'failed : start gateawy%d' % gw_id )
time.sleep(5)
# Check if gateway is added
added_gw = {"ip":HOST_IP,"port":gw_port}
log_reader = LogReader(api.conf.log_file_prefix)
found = False
while True:
line = log_reader.readline()
if line == None:
break
if line.find(MSG_GATEWAY_ADD_ZK) == -1:
continue
gw = line.split('data:')[1]
gw = ast.literal_eval(gw)
if gw['ip'] == added_gw['ip'] and gw['port'] == added_gw['port']:
found = True
if not found:
self.fail('FAIL, load gateway information, gw:%s' % util.json_to_str(added_gw))
else:
util.log('SUCCESS, load gateway information.')
# Delete gateway
cmd = 'gw_del %s %d' % (CLUSTER_NAME, gw_id)
ret = util.cm_command(MGMT_IP, MGMT_PORT, cmd)
if ret != None and len(ret) >= 2:
ret = ret[:-2]
util.log('cmd:"%s", ret:"%s"' % (cmd, ret))
if not ret.startswith('{"state":"success","msg":"+OK"}'):
self.fail('failed to delete gateway')
# Check if gateway is deleted
deleted_gw = {"ip":HOST_IP,"port":gw_port}
found = check_gateway_deleted(deleted_gw, api)
if not found:
self.fail('FAIL, delete gateway information, gw:%s' % util.json_to_str(deleted_gw))
else:
util.log('SUCCESS, delete gateway information.')
# Stop gateway
ret = util.shutdown_gateway(gw_id, gw_port)
self.assertEqual(ret, 0, 'failed : shutdown gateawy%d' % gw_id)
api.destroy()
def test_6_repeat_isolation_and_no_opinion_linepay(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
out = util.sudo('iptables -t nat -A OUTPUT -d 127.0.0.100 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'add a forwarding role to iptables fail. output:%s' % out)
out = util.sudo('iptables -t nat -A PREROUTING -d 127.0.0.100 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'add a forwarding role to iptables fail. output:%s' % out)
out = util.sudo('iptables -t nat -A OUTPUT -d 127.0.0.101 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'add a forwarding role to iptables fail. output:%s' % out)
out = util.sudo('iptables -t nat -A PREROUTING -d 127.0.0.101 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'add a forwarding role to iptables fail. output:%s' % out)
out = util.sudo('iptables -t nat -A OUTPUT -d 127.0.0.102 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'add a forwarding role to iptables fail. output:%s' % out)
out = util.sudo('iptables -t nat -A PREROUTING -d 127.0.0.102 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'add a forwarding role to iptables fail. output:%s' % out)
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
ret = default_cluster.initialize_starting_up_smr_before_redis( cluster )
self.assertEqual(0, ret, 'failed to TestMaintenance.initialize')
# 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)
out = util.sudo('iptables -A OUTPUT -d 127.0.0.102 -j DROP')
self.assertTrue(out.succeeded, 'add a bloking role to iptables fail. output:%s' % out)
for i in range(1):
util.log('waiting... %d' % (i + 1))
time.sleep(0.1)
out = util.sudo('iptables -A OUTPUT -d 127.0.0.100 -j DROP')
self.assertTrue(out.succeeded, 'add a bloking role to iptables fail. output:%s' % out)
for i in range(3):
util.log('waiting... %d' % (i + 1))
time.sleep(1.2)
out = util.sudo('iptables -A OUTPUT -d 127.0.0.101 -j DROP')
self.assertTrue(out.succeeded, 'add a bloking role to iptables fail. output:%s' % out)
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)
out = util.sudo('iptables -D OUTPUT -d 127.0.0.102 -j DROP')
self.assertTrue(out.succeeded, 'delete a bloking role to iptables fail. output:%s' % out)
for i in range(0):
util.log('waiting... %d' % (i + 1))
time.sleep(1)
out = util.sudo('iptables -D OUTPUT -d 127.0.0.100 -j DROP')
self.assertTrue(out.succeeded, 'delete a bloking role to iptables fail. output:%s' % out)
for i in range(0):
util.log('waiting... %d' % (i + 1))
time.sleep(1)
out = util.sudo('iptables -D OUTPUT -d 127.0.0.101 -j DROP')
self.assertTrue(out.succeeded, 'delete a bloking role to iptables fail. output:%s' % out)
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')
# Shutdown cluster
ret = default_cluster.finalize( cluster )
self.assertEqual(ret, 0, 'failed to TestMaintenance.finalize')
# Delete forwarding role
out = util.sudo('iptables -t nat -D OUTPUT -d 127.0.0.100 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'delete a forwarding role to iptables fail. output:%s' % out)
out = util.sudo('iptables -t nat -D PREROUTING -d 127.0.0.100 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'delete a forwarding role to iptables fail. output:%s' % out)
out = util.sudo('iptables -t nat -D OUTPUT -d 127.0.0.101 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'delete a forwarding role to iptables fail. output:%s' % out)
out = util.sudo('iptables -t nat -D PREROUTING -d 127.0.0.101 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'delete a forwarding role to iptables fail. output:%s' % out)
out = util.sudo('iptables -t nat -D OUTPUT -d 127.0.0.102 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'delete a forwarding role to iptables fail. output:%s' % out)
out = util.sudo('iptables -t nat -D PREROUTING -d 127.0.0.102 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'delete a forwarding role to iptables fail. output:%s' % out)
def test_7_dirty_network_fi(self):
util.print_frame()
clnts = []
try:
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
out = util.sudo('iptables -t nat -A OUTPUT -d 127.0.0.100 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'add a forwarding role to iptables fail. output:%s' % out)
out = util.sudo('iptables -t nat -A PREROUTING -d 127.0.0.100 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'add a forwarding role to iptables fail. output:%s' % out)
cluster_name = 'network_isolation_cluster_1'
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
ret = default_cluster.initialize_starting_up_smr_before_redis( cluster,
conf={'cm_context':'applicationContext-fi.xml'})
self.assertEqual(0, ret, 'failed to TestMaintenance.initialize')
# 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')
# Start crc16 client
for s in cluster['servers']:
c = load_generator_crc16.Crc16Client(s['id'], s['ip'], s['gateway_port'], 3000, verbose=False)
c.start()
clnts.append(c)
# Network isolation test
cmfi = fi_confmaster.ConfmasterWfFi(['ra', 'qa', 'me', 'yj', 'bj', 'mg'],
['lconn', 'slave', 'master', 'setquorum'], [True, False], 1)
for fi in cmfi:
# Block network
util.log('\n\n\n ### BLOCK NETWORK, %s ### ' % str(fi))
ret = block_network(cluster, mgmt_ip, mgmt_port)
self.assertTrue(ret, '[%s] failed to block network.' % str(fi))
for i in xrange(4):
util.log('waiting... %d' % (i + 1))
time.sleep(1)
# Check cluster state
ok = False
for i in xrange(10):
isolated_states = []
util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, isolated_states, check_quorum=True)
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')
# Fault injection
try:
self.assertTrue(fi_confmaster.fi_add(fi, 1, mgmt_ip, mgmt_port),
"Confmaster command fail. fi: %s" % str(fi))
except ValueError as e:
self.fail("Confmaster command error. cmd: \"%s\", reply: \"%s\"" % (cmd, reply))
# Unblock network
util.log('\n\n\n ### UNBLOCK NETWORK, %s ### ' % str(fi))
ret = unblock_network(cluster, mgmt_ip, mgmt_port, None)
self.assertTrue(ret, '[%s] failed to unblock network.' % str(fi))
for i in xrange(4):
util.log('waiting... %d' % (i + 1))
time.sleep(1)
# Check cluster state
ok = False
for i in xrange(10):
isolated_states = []
ok = util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, isolated_states, check_quorum=True)
if ok:
break
time.sleep(1)
self.assertTrue(ok, '[%s] Fail. unstable cluster.' % str(fi))
check_cluster = False
# 'bj', 'slave'
if fi[0] == 'bj' and fi[1] == 'slave':
m, s1, s2 = util.get_mss(cluster)
ret = util.role_lconn(s1)
self.assertEqual("+OK\r\n", ret, '[%s] role lconn fail.' % str(fi))
check_cluster = True
# 'me', 'lconn'
elif fi[0] == 'me' and fi[1] == 'lconn':
m, s1, s2 = util.get_mss(cluster)
ret = util.role_lconn(m)
self.assertEqual("+OK\r\n", ret, '[%s] role lconn fail.' % str(fi))
check_cluster = True
# 'qa', 'setquorum'
elif fi[0] == 'qa' and fi[1] == 'setquorum':
m, s1, s2 = util.get_mss(cluster)
# shutdown
ret = testbase.request_to_shutdown_smr(s1)
self.assertEqual(0, ret, '[%s] failed to shutdown smr%d' % (str(fi), s1['id']))
ret = testbase.request_to_shutdown_redis(s1)
self.assertEqual(0, ret, '[%s] failed to shutdown redis%d' % (str(fi), s1['id']))
# Check quorum
q = -1
for q_cnt in xrange(20):
q = util.get_quorum(m)
if q == 1:
break
time.sleep(1)
self.assertEquals(1, q, "[%s] check quorum fail." % str(fi))
# Modify quorum
ret = util.cmd_to_smr_addr(m['ip'], m['smr_mgmt_port'], 'setquorum 0\r\n')
self.assertEqual("+OK\r\n", ret, '[%s] "setquorum 0" fail.' % str(fi))
# Check quorum
q = -1
for q_cnt in xrange(20):
q = util.get_quorum(m)
if q == 1:
break
time.sleep(1)
self.assertEquals(1, q, "[%s] check quorum fail." % str(fi))
# recovery
ret = testbase.request_to_start_smr(s1)
self.assertEqual(0, ret, '[%s] failed to start smr' % str(fi))
ret = testbase.request_to_start_redis(s1, max_try=120)
self.assertEqual(0, ret, '[%s] failed to start redis' % str(fi))
ret = testbase.wait_until_finished_to_set_up_role(s1, 11)
self.assertEqual(0, ret, '[%s] failed to role change. smr_id:%d' % (str(fi), s1['id']))
check_cluster = True
# 'setquorum'
elif fi[1] == 'setquorum':
m, s1, s2 = util.get_mss(cluster)
ret = util.cmd_to_smr_addr(s1['ip'], s1['smr_mgmt_port'], 'fi delay sleep 1 8000\r\n', timeout=20)
self.assertEqual("+OK\r\n", ret, '[%s] "fi delay sleep 1 8000" fail. ret: "%s"' % (str(fi), ret))
check_cluster = True
if check_cluster:
# Check cluster state
ok = False
for i in xrange(20):
isolated_states = []
ok = util.check_cluster(cluster['cluster_name'], mgmt_ip, mgmt_port, isolated_states, check_quorum=True)
if ok:
break
time.sleep(1)
self.assertTrue(ok, '[%s] Fail. unstable cluster.' % str(fi))
# Check fault injection
ok = False
for i in xrange(10):
count = fi_confmaster.fi_count(fi, mgmt_ip, mgmt_port)
if count == 0:
ok = True
break
time.sleep(0.5)
self.assertTrue(ok, "[%s] fail. failt injection had not been triggered." % str(fi))
# Shutdown cluster
ret = default_cluster.finalize( cluster )
self.assertEqual(ret, 0, '[%s] failed to TestMaintenance.finalize' % str(fi))
# Delete forwarding role
out = util.sudo('iptables -t nat -D OUTPUT -d 127.0.0.100 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'delete a forwarding role to iptables fail. output:%s' % out)
out = util.sudo('iptables -t nat -D PREROUTING -d 127.0.0.100 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'delete a forwarding role to iptables fail. output:%s' % out)
for c in clnts:
self.assertTrue(c.is_consistency(), '[%s] data consistency error!' % str(fi))
finally:
for c in clnts:
c.quit()
for c in clnts:
c.join()
def test_5_mgmt_is_isolated_with_master_failover(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)
out = util.sudo('iptables -t nat -A OUTPUT -d 127.0.0.100 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'add a forwarding role to iptables fail. output:%s' % out)
out = util.sudo('iptables -t nat -A PREROUTING -d 127.0.0.100 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'add a forwarding role to iptables fail. output:%s' % out)
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
ret = default_cluster.initialize_starting_up_smr_before_redis( cluster )
self.assertEqual(0, ret, 'failed to TestMaintenance.initialize')
# 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 loop_cnt in range(3):
master, slave1, slave2 = util.get_mss(cluster)
self.assertNotEquals(master, None, 'there is no master')
self.assertNotEquals(slave1, None, 'there is no slave1')
self.assertNotEquals(slave2, None, 'there is no slave2')
# Block network
util.log('\n\n\n ### BLOCK NETWORK, %d ### ' % loop_cnt)
out = util.sudo('iptables -A OUTPUT -d 127.0.0.100 -j DROP')
self.assertTrue(out.succeeded, 'add a bloking role to iptables fail. output:%s' % out)
for i in range(4):
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['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')
# Shutdown master
util.log( 'shutdown pgs%d while hanging.' % master['id'] )
ret = testbase.request_to_shutdown_smr( master )
self.assertEqual( ret, 0, 'failed to shutdown smr. id:%d' % master['id'] )
ret = testbase.request_to_shutdown_redis( master )
self.assertEqual( ret, 0, 'failed to shutdown redis. id:%d' % master['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( master['id'], cluster['cluster_name'], mgmt_ip, mgmt_port )
if expected == state:
break;
time.sleep( 1 )
self.assertEqual( expected , state,
'master%d - state:%s, expected:%s' % (master['id'], state, expected) )
util.log( 'succeeded : pgs%d state changed to F.' % master['id'] )
# Unblock network
util.log('\n\n\n ### UNBLOCK NETWORK, %d ### ' % loop_cnt)
out = util.sudo('iptables -D OUTPUT -d 127.0.0.100 -j DROP')
self.assertTrue(out.succeeded, 'delete a bloking role to iptables fail. output:%s' % out)
# 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'] == master['id']:
continue
if s['active_role'] != s['mgmt_role']:
state_consistency = False
if state_consistency:
ok = True
break
time.sleep(1)
self.assertTrue(ok, 'Fail, state consistency')
# Recovery
util.log( 'restart pgs%d.' % master['id'] )
ret = testbase.request_to_start_smr( master )
self.assertEqual( ret, 0, 'failed to start smr. id:%d' % master['id'] )
ret = testbase.request_to_start_redis( master )
self.assertEqual( ret, 0, 'failed to start redis. id:%d' % master['id'] )
wait_count = 20
ret = testbase.wait_until_finished_to_set_up_role( master, wait_count )
self.assertEqual( ret, 0, 'failed to role change. smr_id:%d' % (master['id']) )
redis = redis_mgmt.Redis( master['id'] )
ret = redis.connect( master['ip'], master['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')
# 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(0, 3):
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)
for i in range(3, 6):
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.assertEqual(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')
# Shutdown cluster
ret = default_cluster.finalize( cluster )
self.assertEqual(ret, 0, 'failed to TestMaintenance.finalize')
# Delete forwarding role (127.0.0.100 -> 127.0.0.1)
out = util.sudo('iptables -t nat -D OUTPUT -d 127.0.0.100 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'delete a forwarding role to iptables fail. output:%s' % out)
out = util.sudo('iptables -t nat -D PREROUTING -d 127.0.0.100 -p tcp -j DNAT --to-destination 127.0.0.1')
self.assertTrue(out.succeeded, 'delete a forwarding role to iptables fail. output:%s' % out)
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 _gateway_fault_failback(self, api):
# Set up arguments
server = self.cluster['servers'][0]
gw_id = server['id']
gw_port = server['gateway_port']
# Stop gateway
ret = util.shutdown_gateway(gw_id, gw_port, True)
self.assertEqual(ret, 0, 'failed : shutdown gateawy%d' % gw_id)
time.sleep(3)
# Check error
saved_err_cnt = {}
for i in range(len(self.load_gen_list)):
saved_err_cnt[i] = 0
no_error_cnt = 0
for i in range(10):
util.log('check error count loop:%d' % i)
no_err = True
for j in range(len(self.load_gen_list)):
err_cnt = self.load_gen_list[j].get_err_cnt()
if err_cnt != saved_err_cnt[j]:
no_err = False
util.log('saved_err_cnt:%d, err_cnt:%d' % (saved_err_cnt[j], err_cnt))
saved_err_cnt[j] = err_cnt
if no_err:
no_error_cnt += 1
if no_error_cnt >= 3:
break
time.sleep(1)
self.assertTrue(no_error_cnt >= 3, 'failed to get replys from well working gateways')
# Check if gateway is deleted
deleted_gw = {"ip":HOST_IP,"port":gw_port}
found = check_gateway_deleted(deleted_gw, api)
if not found:
self.fail('FAIL, delete gateway information, gw:%s' % util.json_to_str(deleted_gw))
else:
util.log('SUCCESS, delete gateway information.')
# Start gateway
ret = util.start_gateway( gw_id, server['ip'], MGMT_PORT, server['cluster_name'], gw_port)
self.assertEqual( ret, 0, 'failed : start gateawy%d' % gw_id )
time.sleep(3)
# Check if gateway is added
found = check_gateway_added(deleted_gw, api)
if not found:
self.fail('FAIL, load gateway information, gw:%s' % util.json_to_str(added_gw))
else:
util.log('SUCCESS, load gateway information.')
# Check loadbalancing
for i in range(3):
ok = True
for s in self.cluster['servers']:
tps = util.get_tps(s['ip'], s['gateway_port'], 'gw')
if tps < 50:
ok = False
if ok:
break
time.sleep(1)
if not ok:
self.fail('FAIL, loadbalancing,')
else:
util.log('SUCCESS, loadbalancing.')
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_gateway_upgrade(self):
util.print_frame()
api = ARC_API(ZK_ADDR, CLUSTER_NAME, logFilePrefix = self.arcci_log, so_path = self.so_path)
# Start load generation
self.load_gen_list = {}
for i in range(len(self.cluster['servers'])):
arc_api = ARC_API(ZK_ADDR, CLUSTER_NAME, logFilePrefix = self.arcci_log, so_path = self.so_path)
server = self.cluster['servers'][i]
load_gen = LoadGenerator_ARCCI(server['id'], arc_api, timeout_second=10)
load_gen.start()
self.load_gen_list[i] = load_gen
# Set up arguments
server = self.cluster['servers'][0]
gw_id = server['id']
gw_port = server['gateway_port']
# Check load
for i in range(5):
ok = True
for s in self.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, 'failed to send requests')
# Delete gateway
self.assertTrue(
util.gw_del(CLUSTER_NAME, gw_id, MGMT_IP, MGMT_PORT),
'failed to delete gateway')
# Check load
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, 'failed to send requests')
# Stop gateway
ret = util.shutdown_gateway(gw_id, gw_port, True)
self.assertEqual(ret, 0, 'failed : shutdown gateawy%d' % gw_id)
# Check if gateway is deleted
deleted_gw = {"ip":HOST_IP,"port":gw_port}
found = check_gateway_deleted(deleted_gw, api)
if not found:
self.fail('FAIL, delete gateway information, gw:%s' % util.json_to_str(deleted_gw))
else:
util.log('SUCCESS, delete gateway information.')
# Start gateway
ret = util.start_gateway( gw_id, server['ip'], MGMT_PORT, server['cluster_name'], gw_port)
self.assertEqual( ret, 0, 'failed : start gateawy%d' % gw_id )
time.sleep(3)
# Add gateway
self.assertTrue(
util.gw_add(CLUSTER_NAME, gw_id, HOST_NAME, HOST_IP, gw_port, MGMT_IP, MGMT_PORT),
'failed to add gateway')
# Check if gateway is added
added_gw = {"ip":HOST_IP,"port":gw_port}
log_reader = LogReader(api.conf.log_file_prefix)
found = False
while True:
line = log_reader.readline()
if line == None:
break
if line.find(MSG_GATEWAY_ADD_ZK) == -1:
continue
gw = line.split('data:')[1]
gw = ast.literal_eval(gw)
if gw['ip'] == added_gw['ip'] and gw['port'] == added_gw['port']:
found = True
if not found:
self.fail('FAIL, load gateway information, gw:%s' % util.json_to_str(added_gw))
else:
util.log('SUCCESS, load gateway information.')
# Check loadbalancing
for i in range(5):
ok = True
for s in self.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)
if not ok:
self.fail('FAIL, loadbalancing,')
else:
util.log('SUCCESS, loadbalancing.')
api.destroy()
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_1_mgmt_is_isolated(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)
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
ret = default_cluster.initialize_starting_up_smr_before_redis( cluster )
self.assertEqual(0, ret, 'failed to TestMaintenance.initialize')
# 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']:
out = util.sudo('iptables -A OUTPUT -d 127.0.0.100 -p tcp --dport %d -j DROP' % (s['smr_mgmt_port']))
"""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(out.succeeded, 'add a bloking role to iptables fail. output:%s' % out)
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']:
out = util.sudo('iptables -D OUTPUT -d 127.0.0.100 -p tcp --dport %d -j DROP' % (s['smr_mgmt_port']))
self.assertTrue(out.succeeded, 'delete a bloking role to iptables fail. output:%s' % out)
# 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['active_role'] != s['mgmt_role']:
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.assertEqual(initial_state[i]['active_ts'], final_state[i]['active_ts'], msg)
# Shutdown cluster
ret = default_cluster.finalize( cluster )
self.assertEqual(ret, 0, 'failed to TestMaintenance.finalize')