def restorable_time_failover_twice(self):
def failover(id):
rpc.call(utils.get_context(), "taskmanager",
{"method": "failover",
"args": {'instance_id':id}})
self.create()
master = utils.get_builtin_instance(self.inst_id)
master_vid = master.virtual_instance_id
rt_1 = utils.get_restorable_time(master_vid)
# case1: failover only
failover1_time = datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S')
failover(master.id)
utils.check_server_status(master_vid, DBInstanceType.MASTER, timeout = 360)
new_slave_id = master.id
utils.check_server_status(new_slave_id, DBInstanceType.STANDBY, timeout = 360)
rt_2 = utils.get_restorable_time(master_vid)
assert rt_2.end > rt_1.end and rt_1.begin <= rt_2.begin, \
("before first failover rt: [%s, %s], after first failover rt: [%s, %s]" % (rt_1.begin, rt_1.end, rt_2.begin, rt_2.end))
# case2: restart & failover
master2 = utils.get_builtin_instance(master_vid)
failover(master2.id)
utils.check_server_status(master_vid, DBInstanceType.MASTER, timeout = 360)
new_slave_id2 = master2.id
utils.check_server_status(new_slave_id2, DBInstanceType.STANDBY, timeout = 360)
rt_3 = utils.get_restorable_time(master_vid)
assert rt_3.begin > failover1_time and rt_3.end > rt_2.end, (rt_3.begin, failover1_time, rt_3.end, rt_2.end)
print rt_1.begin, rt_1.end
print rt_2.begin, rt_2.end
print rt_3.begin, rt_3.end
print "the first failover time: %s" % (failover1_time)
time.sleep(60)
rt_4 = utils.get_restorable_time(master_vid)
print rt_4.begin, rt_4.end
assert rt_4.end > rt_3.end
def restore_to_point_in_time(self):
self.create()
old_master_inst = utils.get_builtin_instance(self.inst_id)
master_vid = old_master_inst.virtual_instance_id
db_count = 100
utils.generate_databases(self.inst_id, count = db_count)
utils.check_generated_databases(self.inst_id, count = db_count)
time.sleep(3)
time1 = datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S')
utils.clear_generated_databases(self.inst_id, count = db_count)
time.sleep(3)
dt2 = datetime.datetime.now()
time2 = dt2.strftime('%Y-%m-%d %H:%M:%S')
while True:
restorable_time = utils.get_restorable_time(self.inst_id)
dt = datetime.datetime.strptime(restorable_time.end, '%Y-%m-%d %H:%M:%S')
if dt > dt2:
break
time.sleep(3)
utils.restore_to_point_in_time_byclient(master_vid, time1)
time.sleep(10) # wait for compute instance appear
new_inst1 = utils.get_restore_instance()
utils.check_server_status(new_inst1, DBInstanceType.SINGLE, timeout=RESTORE_TIME_OUT)
utils.check_generated_databases(new_inst1, count = db_count)
utils.delete_rds_byclient(new_inst1)
utils.check_server_status(new_inst1, expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.SINGLE,
expected_svr_status=utils.ServiceStatuses.DELETED,
deleted=True, timeout=CONF.trove_delete_timeout)
# delete original instance
utils.delete_rds_byclient(master_vid)
utils.check_server_status(self.inst_id, expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.MASTER,
expected_svr_status=utils.ServiceStatuses.DELETED,
deleted=True, timeout=CONF.trove_delete_timeout)
# restore to point in time after deleting
utils.restore_to_point_in_time_byclient(master_vid, time1)
time.sleep(10) # wait for compute instance appear
new_inst2 = utils.get_restore_instance()
utils.check_server_status(new_inst2, DBInstanceType.SINGLE, timeout=RESTORE_TIME_OUT)
utils.check_generated_databases(new_inst2, count = db_count)
utils.delete_rds_byclient(new_inst2)
utils.check_server_status(new_inst2, expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.SINGLE,
expected_svr_status=utils.ServiceStatuses.DELETED,
deleted=True, timeout=CONF.trove_delete_timeout)
def override_to_point_in_time(self):
self.create()
old_master_inst = utils.get_builtin_instance(self.inst_id)
master_vid = old_master_inst.virtual_instance_id
db_count = 100
utils.generate_databases(self.inst_id, count = db_count)
utils.check_generated_databases(self.inst_id, count = db_count)
time.sleep(3)
time1 = datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S')
utils.clear_generated_databases(self.inst_id, count = db_count)
time.sleep(3)
dt2 = datetime.datetime.now()
time2 = dt2.strftime('%Y-%m-%d %H:%M:%S')
while True:
restorable_time = utils.get_restorable_time(self.inst_id)
dt = datetime.datetime.strptime(restorable_time.end, '%Y-%m-%d %H:%M:%S')
if dt > dt2:
break
time.sleep(3)
utils.override_with_backup_byclient(master_vid, None, time1)
utils.check_server_deleted(old_master_inst.id, DBInstanceType.PENDING, timeout = RESTORE_TIME_OUT)
new_master_inst1 = utils.get_builtin_instance(master_vid)
self.inst_id = new_master_inst1.id
utils.check_generated_databases(self.inst_id, count = db_count)
time.sleep(3)
utils.override_with_backup_byclient(master_vid, None, time2)
utils.check_server_deleted(self.inst_id, DBInstanceType.PENDING, timeout = RESTORE_TIME_OUT)
utils.check_generated_databases(master_vid, count = 0)
self.inst_id = utils.get_instance_id(self.group_id, DBInstanceType.MASTER)
self.dbslave_id = utils.get_instance_id(self.group_id, DBInstanceType.STANDBY)
self.validate()
def _failover_test(self, group_id, trigger_inst_id,
do_workload = False, do_prepare = False,
mysqld_killed = False,
host_rebooted = False,
remove_tmp_initsql = False,
mysql_data_lost = False,
check_vip = False,
check_rpl_consist = True,
check_binlog_range = False):
LOG.info("Doing Failover Test, group_id:%s, instance_id:%s, do_workload:%s, do_prepare:%s." %
(group_id, trigger_inst_id, do_workload, do_prepare))
before_group_items = InstanceGroupItem.list_by_gid(test_utils.get_context(), group_id, deleted = False)
before_items = set(map(lambda x: x.type + "_" + x.instance_id, before_group_items))
before_instance = test_utils.get_builtin_instance( trigger_inst_id)
before_rip = test_utils.check_allocate_ip(before_instance.server)
before_origin_instid = before_instance.id
rt_before = rt_after = None
if check_binlog_range:
rt_before = test_utils.get_restorable_time(trigger_inst_id)
if do_workload and before_instance.type == DBInstanceType.MASTER:
FAILOVERInstance.__run_workload(do_prepare = do_prepare)
if remove_tmp_initsql:
FAILOVERInstance.__trigger_vm_remove_tmp_sql_file(trigger_inst_id)
if mysqld_killed:
FAILOVERInstance.__trigger_mysqld_crash(trigger_inst_id)
test_utils.check_server_status(trigger_inst_id, expected_task=tasks.InstanceTasks.NONE,
type=before_instance.type, expected_svr_status=test_utils.ServiceStatuses.SHUTDOWN,
deleted=False, timeout=120)
if host_rebooted:
FAILOVERInstance.__trigger_host_reboot(trigger_inst_id)
# when host-machine rebooted, no guestagent update service's status.
# test_utils.check_server_status(trigger_inst_id, expected_task=tasks.InstanceTasks.NONE,
# type=before_instance.type, expected_svr_status=test_utils.ServiceStatuses.SHUTDOWN,
# deleted=False, timeout=120)
if mysql_data_lost:
FAILOVERInstance.__trigger_mysql_data_lost(trigger_inst_id)
rpc.call(test_utils.get_context(), "taskmanager",
{"method": "failover", "args": {'instance_id':before_origin_instid}}, timeout = 3600)
## check vip <--> rip mapping.
## vip should be changed in 10 seconds.
if before_instance.type == DBInstanceType.MASTER or before_instance.type == DBInstanceType.READ_REPLI:
after_instance = test_utils.get_builtin_instance( trigger_inst_id)
after_nova_inst = after_instance.server
after_rip = test_utils.check_allocate_ip(after_nova_inst)
assert after_instance.vip == before_instance.vip and before_rip != after_rip
if before_instance.type == DBInstanceType.MASTER:
test_utils.check_server_status(before_instance.id,
expected_task = tasks.InstanceTasks.NONE,
type=DBInstanceType.MASTER,
expected_svr_status = test_utils.ServiceStatuses.RUNNING,
deleted=False, timeout=120)
## check replication topo
after_group_items = InstanceGroupItem.list_by_gid(test_utils.get_context(), group_id, deleted = False)
after_items = set(map(lambda x: x.type + "_" + x.instance_id, after_group_items))
LOG.info("before " + str(before_items))
LOG.info("after " + str(after_items))
if check_rpl_consist:
diff_items = (before_items - after_items)
# assert len(diff_items) == 0
assert len(before_group_items) == len(after_group_items), "size of mysql cluster should be the same."
for group_item in after_group_items:
if group_item.type == DBInstanceType.STANDBY and group_item.instance_id == before_instance.id:
item = InstanceGroupItem.get_by_instance_id(test_utils.get_context(), group_item.instance_id, deleted = False)
assert item != None
continue
test_utils.check_server_status(group_item.instance_id,
expected_task = tasks.InstanceTasks.NONE,
type = group_item.type,
expected_svr_status = test_utils.ServiceStatuses.RUNNING,
deleted = False, timeout = 120)
if check_binlog_range:
rt_after = test_utils.get_restorable_time(trigger_inst_id)
assert rt_after.end > rt_before.end, (rt_after.end, rt_before.end)
time.sleep(60)
rt_after2 = test_utils.get_restorable_time(trigger_inst_id)
assert rt_after2.end > rt_after.end, (rt_after2.end, rt_after.end)
def _get_restorable_time(sleep = None):
if sleep != None:
time.sleep(sleep)
restorable_time = utils.get_restorable_time(master_vid)
return restorable_time