def __check_config_is_right(self):
config = KSC_Configuration.load(utils.get_context(), self.config.id)
if not config:
raise Exception("create config, not write in db")
if self.name != config.name or self.description != config.description:
raise Exception("create config, name or description bad value")
if self.config.config_type != '1':
raise Exception("create config, config_type bad value")
overrides = KSC_Configuration.get_configuration_overrides(utils.get_context(), self.config.id)
diffs = {}
for k, v in self.values.iteritems():
if overrides[k] != v and float(overrides[k]) != float(v):
diffs[k] = (v, overrides[k])
if len(diffs):
raise Exception("create config, values error")
#
# if overrides != self.values:
# raise Exception("create config, values error")
datastore_version = DatastoreVersion.load_by_uuid(self.config.datastore_version_id)
if not datastore_version:
raise Exception("create config, bad value of datastore_version_id")
if datastore_version.name != self.datastore_version:
raise Exception("create config, bad value of datastore_version_id")
def resize_ha(self, instance_id=None):
self.inst_id = instance_id
if not self.inst_id:
self.inst_id = utils.get_instance_id_bytenant(DBInstanceType.MASTER)
_inst = utils.get_instance(self.inst_id, deleted=0)
self.group_id = _inst.group_id
master = InstanceGroupItem.get_by_gid_type(utils.get_context(), self.group_id, DBInstanceType.MASTER)
standby = InstanceGroupItem.get_by_gid_type(utils.get_context(), self.group_id, DBInstanceType.STANDBY)
if master:
old_master_inst_id = master.instance_id
if standby:
old_standy_inst_id = standby.instance_id
old_master_inst = utils.get_builtin_instance(old_master_inst_id)
old_flavor_id = old_master_inst.flavor_id
old_virtual_instance_id = old_master_inst.virtual_instance_id
self.inst_id = old_master_inst.id
if old_flavor_id not in ['1', '2', '3', '4']:
raise Exception("It is not support to do resizing based on flavor id: %s, supported flavor_ids should be in (1,2,3,4)" % (old_flavor_id))
flavor = str(int(old_flavor_id) + 1)
LOG.info("old flavor : %s, new flavor : %s" % (old_flavor_id, flavor))
utils.resize_flavor_byclient(self.inst_id, flavor)
inst_ids = utils.check_resize_status(self.group_id)
if len(inst_ids) >= 2:
utils.check_rpl_delay(inst_ids[0])
utils.check_rpl_delay(inst_ids[1])
import time
time.sleep(60)
new_master = utils.get_builtin_instance(old_virtual_instance_id)
new_standby = InstanceGroupItem.get_by_gid_type(utils.get_context(), new_master.group_id, DBInstanceType.STANDBY)
if new_master.virtual_instance_id == old_virtual_instance_id and \
new_master.id != old_master_inst.id:
self.inst_id = new_master.id
self.dbslave_id = new_standby.instance_id
self.vip_id = utils.get_vip_id(self.inst_id)
else:
raise Exception("resize for ha failed,new_master.virtual_instance_id %s,"
" old_virtual_instance_id %s, new_master.id %s,"
" old_master_inst.id %s, new_standby_id %s,"
" old_standy_inst_id %s" % (
new_master.virtual_instance_id, old_virtual_instance_id,
new_master.id, old_master_inst.id,
new_standby.instance_id, old_standy_inst_id))
self.validate()
else:
raise Exception("the num of instance_id should be equal or greater than two after resize HA ")
def update(self):
self.create()
configuration_id = self.config.id
values = {"myisam_sort_buffer_size":115343360,"join_buffer_size":307200,'sort_buffer_size':327680}
valuesJson = json.dumps(values)
name = 'update_%s' % self.config.name
description = 'update_%s' % self.config.description
utils.configuration_update_byclient(configuration_id, valuesJson, name, description)
config = KSC_Configuration.load(utils.get_context(), configuration_id)
if config == self.config:
raise Exception("update config, something is not update, config = %s" % config)
override = KSC_Configuration.get_configuration_overrides(utils.get_context(), configuration_id)
if values == override:
raise Exception("update config, something is not update, config override = %s" % override)
def migrate(self, instance_id=None):
self.inst_id = instance_id
if not self.inst_id:
self.inst_id = utils.get_instance_id_bytenant(DBInstanceType.SINGLE)
utils.check_server_status(self.inst_id,
expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.SINGLE,
expected_svr_status=utils.ServiceStatuses.RUNNING,
deleted=False)
inst_db_info = utils.get_instance(self.inst_id, deleted= False)
old_nova_server_id = inst_db_info.compute_instance_id
ran_count = 45
utils.generate_databases(self.inst_id, count = ran_count)
_ret = rpc.call(utils.get_context(),
"taskmanager",
{"method": "migrate",
"args": {'migrate_id': inst_db_info.id}})
utils.check_server_status(self.inst_id,
expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.SINGLE,
expected_svr_status=utils.ServiceStatuses.RUNNING,
deleted=False)
new_nova_server_id = utils.get_instance(self.inst_id, deleted= False).compute_instance_id
assert old_nova_server_id != new_nova_server_id
utils.check_generated_databases(self.inst_id, count = ran_count)
self.validate()
def patch(self):
if not self.config:
self.create()
configuration_id = self.config.id
values = json.dumps({"myisam_sort_buffer_size":116391936,"join_buffer_size":409600})
utils.configuration_patch_byclient(configuration_id, values)
override = KSC_Configuration.get_configuration_overrides(utils.get_context(), configuration_id)
if values == override:
raise Exception("patch config, something is not update, override = %s" % override)
def migrate_when_backup_fail(self):
self.inst_id = self._get_rid()
utils.check_server_status(self.inst_id, type = DBInstanceType.READ_REPLI)
backup = utils.create_backup_byclient(self.inst_id)
utils.check_backup_status(backup.id)
bk_info = DBBackup.find_by(utils.get_context(), id = backup.id)
bk_info.state = 'FAILED'
bk_info.save()
self.migrate()
def delete(self):
self.create()
#config_type = "1"
#config = utils.get_config_bytenant(config_type)
config = self.config
utils.configuration_delete_byclient(config.id)
is_OK = 0
#check ...
try:
config = KSC_Configuration.load(utils.get_context(), config.id)
except Exception as e:
is_OK = 1
try:
override = KSC_Configuration.get_configuration_overrides(utils.get_context(), config.id)
except Exception as e1:
is_OK = is_OK + 1
if is_OK != 2:
raise Exception("delete config, no delete success..")
def patch_new(self):
if not self.config:
self.create_new()
configuration_id = self.config.id
values = {"myisam_sort_buffer_size":'xxx',"join_buffer_size":'xxx'}
# values = json.dumps({"myisam_sort_buffer_size":116391936,"join_buffer_size":409600})
utils.configuration_patch_byclient(configuration_id, json.dumps(values))
override = KSC_Configuration.get_configuration_overrides(utils.get_context(), configuration_id)
for k, v in values.iteritems():
if override.get(k) != v:
# if values == override:
raise Exception("patch config, something is not update, override = %s" % override)
def migrate(self):
rr_inst_id = self._get_rid()
old_rr_inst = utils.get_builtin_instance(rr_inst_id)
old_rr_id = old_rr_inst.id
rr_vid = old_rr_inst.virtual_instance_id
_ret = rpc.call(utils.get_context(), "taskmanager",
{"method": "migrate",
"args": {'migrate_id': old_rr_inst.id}})
utils.check_server_status(rr_vid, DBInstanceType.READ_REPLI, InstanceTasks.NONE, utils.ServiceStatuses.RUNNING, timeout = 600)
new_rr_inst = utils.get_builtin_instance(rr_vid)
assert new_rr_inst.virtual_instance_id == rr_vid and new_rr_inst.id != old_rr_id, (rr_vid, new_rr_inst.id)
master_id = utils.get_instance_id_bytenant(DBInstanceType.MASTER)
self._validate(master_id, rr_vid)
def force_migrate(self):
force_host = "rds_zone_1:rds-control-18-220.ksc.com"
self.inst_id = utils.get_instance_id_bytenant(DBInstanceType.SINGLE)
_ret = rpc.call(utils.get_context(),
"taskmanager",
{"method": "migrate",
"args": {'migrate_id': self.inst_id, 'host' : force_host }})
self.inst_id = _ret['id']
self.validate()
new_inst = utils.get_instance(self.inst_id)
nova_id = new_inst.compute_instance_id
info = utils.get_nova_server_info(nova_id)
if info['OS-EXT-SRV-ATTR:hypervisor_hostname'] != force_host:
LOG.error("expected host is %s, finally host is %s" % (force_host, info['OS-EXT-SRV-ATTR:hypervisor_hostname']))
def _failover(self, stop_mysqld = False, rm_mysql_data = False):
self.inst_id = self._get_rid()
utils.check_server_status(self.inst_id,expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.READ_REPLI,
expected_svr_status=utils.ServiceStatuses.RUNNING,
deleted=False,timeout=10)
self.vip_id = utils.get_vip_id(self.inst_id)
_rr_server = utils.get_builtin_instance(self.inst_id)
nova_instance = _rr_server.server
self.group_id = _rr_server.db_info.group_id
instance_id = self.inst_id
ran_count = 56
utils.generate_databases(self.master_id, count = ran_count)
ip = utils.check_allocate_ip(nova_instance)
if stop_mysqld:
utils.stop_mysqld(ip)
if rm_mysql_data:
utils.mysql_data_lost(ip)
utils.check_server_status(self.inst_id,expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.READ_REPLI,
expected_svr_status=utils.ServiceStatuses.SHUTDOWN,
deleted=False,timeout=120)
_ret = rpc.call(utils.get_context(),"taskmanager",
{
"method": "failover",
"args": {'instance_id':instance_id}
}
)
utils.check_server_status(self.inst_id,expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.READ_REPLI,
expected_svr_status=utils.ServiceStatuses.RUNNING,
deleted=False,timeout=120)
self.master_id = utils.get_instance_id(self.group_id,DBInstanceType.MASTER)
self.inst_id = self.inst_id
utils.check_generated_databases(self.inst_id, count = ran_count)
self.validate()
def validate(self):
utils.check_server_status(self.inst_id, expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.MASTER,
expected_svr_status=utils.ServiceStatuses.RUNNING,
deleted=False, timeout=600)
utils.check_server_status(self.dbslave_id, expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.STANDBY,
expected_svr_status=utils.ServiceStatuses.RUNNING,
deleted=False, timeout=600)
rr_items = InstanceGroupItem.get_by_gid_type(utils.get_context(), self.group_id, DBInstanceType.READ_REPLI, deleted = False)
slave_ids = []
for rr in rr_items:
slave_ids.append(rr.instance_id)
slave_ids.append(self.dbslave_id)
utils.check_mysql_adminuser(self.inst_id)
utils.check_mysql_adminuser(self.dbslave_id)
for _id in [self.inst_id, self.dbslave_id]:
utils.check_mysql_is_running(self.inst_id)
utils.check_vip(self.inst_id, vip_id=self.vip_id)
self.backup_id = utils.check_backup(self.group_id)
utils.check_backup_status(self.backup_id)
utils.check_backup_path(self.backup_id)
for slave_id in slave_ids:
utils.check_rpl_delay(slave_id)
master_inst = utils.get_builtin_instance(self.inst_id)
slave_inst = utils.get_builtin_instance(self.dbslave_id)
master_ip = utils.check_allocate_ip(master_inst.server)
slave_ip = utils.check_allocate_ip(slave_inst.server)
LOG.info("master_ip:%s slave_ip:%s" % (master_ip, slave_ip))
utils.check_rpl_consist(self.inst_id, slave_ids, master_ip, [slave_ip])
utils.check_rpl_topo_ha(self.group_id)
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 failover(id):
rpc.call(utils.get_context(), "taskmanager",
{"method": "failover",
"args": {'instance_id':id}})
def failover(self, instance_id=None, _strategy=None):
self.inst_id = instance_id
if not self.inst_id:
self.inst_id = utils.get_instance_id_bytenant(DBInstanceType.MASTER)
_inst = utils.get_builtin_instance(self.inst_id)
self.group_id = _inst.group_id
self.dbslave_id = utils.get_instance_id(self.group_id, DBInstanceType.STANDBY)
utils.check_server_status(self.inst_id, expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.MASTER,
expected_svr_status=utils.ServiceStatuses.RUNNING,
deleted=False, timeout=120)
utils.check_server_status(self.dbslave_id, expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.STANDBY,
expected_svr_status=utils.ServiceStatuses.RUNNING,
deleted=False, timeout=120)
self.vip_id = utils.get_vip_id(self.inst_id)
strategy = CONF.ha_failover_strategy
virtual_instance_id = None
if strategy == 'master':
_ret = utils.get_builtin_instance(self.inst_id)
nova_instance = _ret.server
instance_id = _ret.id
type = DBInstanceType.MASTER
virtual_instance_id = _ret.virtual_instance_id
elif strategy == 'standby':
_ret = utils.get_builtin_instance(self.dbslave_id)
nova_instance = _ret.server
instance_id = _ret.id
type = DBInstanceType.STANDBY
else:
raise Exception("not found ha_failover_strategy %s" % strategy)
rancount = random.randint(50, 100)
utils.generate_databases(self.inst_id, count = rancount)
ip = utils.check_allocate_ip(nova_instance)
utils.stop_mysqld(ip, stop_ga=True)
utils.check_server_status(instance_id, expected_task=utils.tasks.InstanceTasks.NONE,
type=type,
expected_svr_status=utils.ServiceStatuses.SHUTDOWN,
deleted=False, timeout=120)
rpc.call(utils.get_context(), "taskmanager",
{"method": "failover",
"args": {'instance_id':instance_id}})
if strategy == 'master':
origin_inst_id = inst_utils.virtual_instid_2_origin_instid(virtual_instance_id)
self.inst_id = origin_inst_id
utils.check_server_status(origin_inst_id, expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.MASTER,
expected_svr_status=utils.ServiceStatuses.RUNNING,
deleted=False, timeout=120)
new_slave_id = utils.get_instance_id(self.group_id, DBInstanceType.STANDBY)
utils.check_server_status(new_slave_id, expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.STANDBY,
expected_svr_status=utils.ServiceStatuses.RUNNING,
deleted=False, timeout=120)
self.dbslave_id = new_slave_id
utils.check_generated_databases(self.inst_id, count = rancount)
utils.clear_generated_databases(self.dbslave_id, count = rancount)
self.validate()
def migrate(self, instance_id=None, _strategy=None):
self.inst_id = instance_id
if not self.inst_id:
self.inst_id = utils.get_instance_id_bytenant(DBInstanceType.MASTER)
_master_inst = utils.get_instance(id = self.inst_id, deleted = 0)
self.group_id = _master_inst.group_id
self.dbslave_id = utils.get_instance_id(self.group_id, DBInstanceType.STANDBY)
utils.check_server_status(self.inst_id, expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.MASTER,
expected_svr_status=utils.ServiceStatuses.RUNNING,
deleted=False, timeout=120)
utils.check_server_status(self.dbslave_id, expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.STANDBY,
expected_svr_status=utils.ServiceStatuses.RUNNING,
deleted=False, timeout=120)
self.vip_id = utils.get_vip_id(_master_inst.id)
virtual_instance_id = None
strategy = CONF.migrate_strategy
if strategy == 'master':
_ret = utils.get_builtin_instance(_master_inst.id)
nova_instance = _ret.server
instance_id = _master_inst.id
type = DBInstanceType.MASTER
virtual_instance_id = _ret.virtual_instance_id
elif strategy == 'standby':
_ret = utils.get_builtin_instance(self.dbslave_id)
nova_instance = _ret.server
instance_id = self.dbslave_id
type = DBInstanceType.STANDBY
else:
raise Exception("not found migrate_strategy ss%s" % strategy)
ran_count = random.randint(50, 100)
utils.generate_databases(self.inst_id, count = ran_count)
_ret = rpc.call(utils.get_context(), "taskmanager",
{"method": "migrate",
"args": {'migrate_id':instance_id}})
if strategy == 'master':
raw_instance_id = inst_utils.virtual_instid_2_origin_instid(virtual_instance_id)
new_server_id = utils.get_builtin_instance(raw_instance_id).server_id
utils.check_server_status(raw_instance_id, expected_task=utils.tasks.InstanceTasks.NONE,
type=DBInstanceType.MASTER,
expected_svr_status=utils.ServiceStatuses.RUNNING,
deleted=False, timeout=120)
assert new_server_id != nova_instance.id
self.inst_id = raw_instance_id
elif strategy == "standby":
utils.check_server_status(_ret['id'], expected_task=utils.tasks.InstanceTasks.NONE,
type = DBInstanceType.STANDBY,
expected_svr_status=utils.ServiceStatuses.RUNNING,
deleted=False, timeout = 123)
self.dbslave_id = _ret['id']
utils.check_generated_databases(self.inst_id, count = ran_count)
utils.check_generated_databases(self.dbslave_id, count = ran_count)
utils.clear_generated_databases(self.inst_id, count = ran_count)
self.validate()
