class ZKClient(object):
def __init__(self,
zk_hosts,
local_port,
local_host=None,
server_info_path='/crawlers'):
self.zk_hosts = zk_hosts
self.local_host = local_host
if not self.local_host:
self.local_host = get_local_host()
self.server_info_path = server_info_path
self.server_info_znode = '{server_info_path}/{ip}:{port}'.format(
server_info_path=server_info_path,
ip=self.local_host,
port=local_port)
self.connect_zk()
def connect_zk(self):
self.zk = KazooClient(hosts=self.zk_hosts)
self.zk.add_listener(self.state_listener)
self.zk.start()
def update_heartbeat(self):
def callback(async_stat):
stat = async_stat.get()
if stat:
_ = self.zk.set_async(self.server_info_znode, ts)
else:
_ = self.zk.create_async(self.server_info_znode,
ts,
ephemeral=True,
makepath=True)
ts = str(int(time.time()))
async_stat = self.zk.exists_async(self.server_info_znode, watch=None)
async_stat.rawlink(callback)
def state_listener(self, state):
if state == KazooState.LOST:
# Register somewhere that the session was lost
pass
elif state == KazooState.SUSPENDED:
# Handle being disconnected from Zookeeper
pass
else:
self.update_heartbeat()
def add_watcher(self, request_handler):
@self.zk.DataWatch(ZOO_CONFIG_PROXY_PATH)
def proxy_change(data, stat):
run_in_thread(request_handler.config_proxy_via_zookeeper, data)
def close(self):
self.zk.stop()
class ZooKeeper(AbstractDCS):
def __init__(self, config):
super(ZooKeeper, self).__init__(config)
hosts = config.get('hosts', [])
if isinstance(hosts, list):
hosts = ','.join(hosts)
self._client = KazooClient(hosts, handler=PatroniSequentialThreadingHandler(config['retry_timeout']),
timeout=config['ttl'], connection_retry={'max_delay': 1, 'max_tries': -1},
command_retry={'deadline': config['retry_timeout'], 'max_delay': 1, 'max_tries': -1})
self._client.add_listener(self.session_listener)
self._my_member_data = None
self._fetch_cluster = True
self._last_leader_operation = 0
self._orig_kazoo_connect = self._client._connection._connect
self._client._connection._connect = self._kazoo_connect
self._client.start()
def _kazoo_connect(self, host, port):
"""Kazoo is using Ping's to determine health of connection to zookeeper. If there is no
response on Ping after Ping interval (1/2 from read_timeout) it will consider current
connection dead and try to connect to another node. Without this "magic" it was taking
up to 2/3 from session timeout (ttl) to figure out that connection was dead and we had
only small time for reconnect and retry.
This method is needed to return different value of read_timeout, which is not calculated
from negotiated session timeout but from value of `loop_wait`. And it is 2 sec smaller
than loop_wait, because we can spend up to 2 seconds when calling `touch_member()` and
`write_leader_optime()` methods, which also may hang..."""
ret = self._orig_kazoo_connect(host, port)
return max(self.loop_wait - 2, 2)*1000, ret[1]
def session_listener(self, state):
if state in [KazooState.SUSPENDED, KazooState.LOST]:
self.cluster_watcher(None)
def cluster_watcher(self, event):
self._fetch_cluster = True
self.event.set()
def reload_config(self, config):
self.set_retry_timeout(config['retry_timeout'])
loop_wait = config['loop_wait']
loop_wait_changed = self._loop_wait != loop_wait
self._loop_wait = loop_wait
self._client.handler.set_connect_timeout(loop_wait)
# We need to reestablish connection to zookeeper if we want to change
# read_timeout (and Ping interval respectively), because read_timeout
# is calculated in `_kazoo_connect` method. If we are changing ttl at
# the same time, set_ttl method will reestablish connection and return
# `!True`, otherwise we will close existing connection and let kazoo
# open the new one.
if not self.set_ttl(int(config['ttl'] * 1000)) and loop_wait_changed:
self._client._connection._socket.close()
def set_ttl(self, ttl):
"""It is not possible to change ttl (session_timeout) in zookeeper without
destroying old session and creating the new one. This method returns `!True`
if session_timeout has been changed (`restart()` has been called)."""
if self._client._session_timeout != ttl:
self._client._session_timeout = ttl
self._client.restart()
return True
def set_retry_timeout(self, retry_timeout):
self._client._retry.deadline = retry_timeout
def get_node(self, key, watch=None):
try:
ret = self._client.get(key, watch)
return (ret[0].decode('utf-8'), ret[1])
except NoNodeError:
return None
@staticmethod
def member(name, value, znode):
return Member.from_node(znode.version, name, znode.ephemeralOwner, value)
def get_children(self, key, watch=None):
try:
return self._client.get_children(key, watch)
except NoNodeError:
return []
def load_members(self):
members = []
for member in self.get_children(self.members_path, self.cluster_watcher):
data = self.get_node(self.members_path + member)
if data is not None:
members.append(self.member(member, *data))
return members
def _inner_load_cluster(self):
self._fetch_cluster = False
self.event.clear()
nodes = set(self.get_children(self.client_path(''), self.cluster_watcher))
if not nodes:
self._fetch_cluster = True
# get initialize flag
initialize = (self.get_node(self.initialize_path) or [None])[0] if self._INITIALIZE in nodes else None
# get global dynamic configuration
config = self.get_node(self.config_path, watch=self.cluster_watcher) if self._CONFIG in nodes else None
config = config and ClusterConfig.from_node(config[1].version, config[0], config[1].mzxid)
# get list of members
members = self.load_members() if self._MEMBERS[:-1] in nodes else []
# get leader
leader = self.get_node(self.leader_path) if self._LEADER in nodes else None
if leader:
client_id = self._client.client_id
if leader[0] == self._name and client_id is not None and client_id[0] != leader[1].ephemeralOwner:
logger.info('I am leader but not owner of the session. Removing leader node')
self._client.delete(self.leader_path)
leader = None
if leader:
member = Member(-1, leader[0], None, {})
member = ([m for m in members if m.name == leader[0]] or [member])[0]
leader = Leader(leader[1].version, leader[1].ephemeralOwner, member)
self._fetch_cluster = member.index == -1
# failover key
failover = self.get_node(self.failover_path, watch=self.cluster_watcher) if self._FAILOVER in nodes else None
failover = failover and Failover.from_node(failover[1].version, failover[0])
# get last leader operation
optime = self.get_node(self.leader_optime_path) if self._OPTIME in nodes and self._fetch_cluster else None
self._last_leader_operation = 0 if optime is None else int(optime[0])
self._cluster = Cluster(initialize, config, leader, self._last_leader_operation, members, failover)
def _load_cluster(self):
if self._fetch_cluster or self._cluster is None:
try:
self._client.retry(self._inner_load_cluster)
except Exception:
logger.exception('get_cluster')
self.cluster_watcher(None)
raise ZooKeeperError('ZooKeeper in not responding properly')
def _create(self, path, value, **kwargs):
try:
self._client.retry(self._client.create, path, value.encode('utf-8'), **kwargs)
return True
except:
return False
def attempt_to_acquire_leader(self, permanent=False):
ret = self._create(self.leader_path, self._name, makepath=True, ephemeral=not permanent)
if not ret:
logger.info('Could not take out TTL lock')
return ret
def set_failover_value(self, value, index=None):
try:
self._client.retry(self._client.set, self.failover_path, value.encode('utf-8'), version=index or -1)
return True
except NoNodeError:
return value == '' or (index is None and self._create(self.failover_path, value))
except:
logging.exception('set_failover_value')
return False
def set_config_value(self, value, index=None):
try:
self._client.retry(self._client.set, self.config_path, value.encode('utf-8'), version=index or -1)
return True
except NoNodeError:
return index is None and self._create(self.config_path, value)
except Exception:
logging.exception('set_config_value')
return False
def initialize(self, create_new=True, sysid=""):
return self._create(self.initialize_path, sysid, makepath=True) if create_new \
else self._client.retry(self._client.set, self.initialize_path, sysid.encode("utf-8"))
def touch_member(self, data, ttl=None, permanent=False):
cluster = self.cluster
member = cluster and ([m for m in cluster.members if m.name == self._name] or [None])[0]
data = data.encode('utf-8')
if member and self._client.client_id is not None and member.session != self._client.client_id[0]:
try:
self._client.delete_async(self.member_path).get(timeout=1)
except NoNodeError:
pass
except:
return False
member = None
if member:
if data == self._my_member_data:
return True
else:
try:
self._client.create_async(self.member_path, data, makepath=True, ephemeral=not permanent).get(timeout=1)
self._my_member_data = data
return True
except Exception as e:
if not isinstance(e, NodeExistsError):
logger.exception('touch_member')
return False
try:
self._client.set_async(self.member_path, data).get(timeout=1)
self._my_member_data = data
return True
except:
logger.exception('touch_member')
return False
def take_leader(self):
return self.attempt_to_acquire_leader()
def write_leader_optime(self, last_operation):
last_operation = last_operation.encode('utf-8')
if last_operation != self._last_leader_operation:
try:
self._client.set_async(self.leader_optime_path, last_operation).get(timeout=1)
self._last_leader_operation = last_operation
except NoNodeError:
try:
self._client.create_async(self.leader_optime_path, last_operation, makepath=True).get(timeout=1)
self._last_leader_operation = last_operation
except:
logger.exception('Failed to create %s', self.leader_optime_path)
except:
logger.exception('Failed to update %s', self.leader_optime_path)
def update_leader(self):
return True
def delete_leader(self):
self._client.restart()
self._my_member_data = None
return True
def _cancel_initialization(self):
node = self.get_node(self.initialize_path)
if node:
self._client.delete(self.initialize_path, version=node[1].version)
def cancel_initialization(self):
try:
self._client.retry(self._cancel_initialization)
except:
logger.exception("Unable to delete initialize key")
def delete_cluster(self):
try:
return self._client.retry(self._client.delete, self.client_path(''), recursive=True)
except NoNodeError:
return True
def watch(self, timeout):
if super(ZooKeeper, self).watch(timeout):
self._fetch_cluster = True
return self._fetch_cluster
class ZookeeperServiceRegistry(BaseServiceRegistry):
def __init__(self, hosts=DEFAULT_HOSTS, chroot=DEFAULT_CHROOT):
super(ZookeeperServiceRegistry, self).__init__()
self.chroot = chroot
self.client = KazooClient(
hosts=hosts,
handler=SequentialGeventHandler(),
)
self.client.add_listener(self.on_kazoo_state_change)
self.start_count = 0
@classmethod
def from_config(cls, config, **kwargs):
return cls(hosts=config.get('hosts', DEFAULT_HOSTS),
chroot=config.get('chroot', DEFAULT_CHROOT),
**kwargs)
def on_start(self, timeout=10):
self.start_count += 1
if self.start_count > 1:
return
started = self.client.start_async()
started.wait(timeout=timeout)
if not self.client.connected:
raise RuntimeError('could not connect to zookeeper')
logger.debug('connected to zookeeper (version=%s)',
'.'.join(map(str, self.client.server_version())))
def on_stop(self):
self.start_count -= 1
if self.start_count != 0:
return
self.client.stop()
def on_kazoo_state_change(self, state):
logger.info('kazoo connection state changed to %s', state)
def on_service_type_watch(self, service, event):
try:
if event.type == EventType.CHILD:
# FIXME: figure out proper retry strategy
self.client.retry(self.lookup, service.container, service)
except Exception:
logger.exception('error in service type watcher')
def on_service_watch(self, service, event):
try:
prefix, service_type, identity = event.path.rsplit('/', 2)
if event.type == EventType.DELETED:
service.remove(identity)
except Exception:
logger.exception('error in service watcher')
def _get_service_znode(self, service, service_type, identity):
path = self._get_zk_path(service_type, identity)
result = self.client.get_async(path,
watch=functools.partial(
self.on_service_watch, service))
value, znode = result.get()
items = six.iteritems(json.loads(value.decode('utf-8')))
return {str(k): str(v) for k, v in items}
def discover(self, container):
result = self.client.get_children_async(path='%s/services' %
self.chroot, )
return list(result.get())
def lookup(self, container, service, watch=True, timeout=1):
def child_watch(event):
print(event)
service_type = service.service_type
result = self.client.get_children_async(
path='%s/services/%s' % (self.chroot, service_type),
watch=functools.partial(self.on_service_type_watch, service),
)
try:
names = result.get(timeout=timeout)
except NoNodeError:
raise LookupFailure(None,
"failed to resolve %s" % service.service_type)
logger.info("lookup %s %r", service_type, names)
identities = set(service.identities())
for name in names:
kwargs = self._get_service_znode(service, service_type, name)
identity = kwargs.pop('identity')
service.update(identity, **kwargs)
try:
identities.remove(identity)
except KeyError:
pass
for identity in identities:
service.remove(identity)
return service
def _get_zk_path(self, service_type, identity):
return '%s/services/%s/%s' % (self.chroot, service_type, identity)
def register(self, container, service_type, timeout=1):
path = self._get_zk_path(service_type, container.identity)
value = json.dumps({
'endpoint': container.endpoint,
'identity': container.identity,
'log_endpoint': container.log_endpoint,
})
result = self.client.create_async(path,
value.encode('utf-8'),
ephemeral=True,
makepath=True)
# FIXME: result.set_exception(RegistrationFailure())
result.get(timeout=timeout)
def unregister(self, container, service_type, timeout=1):
path = self._get_zk_path(service_type, container.identity)
result = self.client.delete_async(path)
result.set_exception(RegistrationFailure())
result.get(timeout=timeout)
class ZooKeeper(AbstractDCS):
def __init__(self, config):
super(ZooKeeper, self).__init__(config)
hosts = config.get('hosts', [])
if isinstance(hosts, list):
hosts = ','.join(hosts)
mapping = {
'use_ssl': 'use_ssl',
'verify': 'verify_certs',
'cacert': 'ca',
'cert': 'certfile',
'key': 'keyfile',
'key_password': '******'
}
kwargs = {v: config[k] for k, v in mapping.items() if k in config}
self._client = KazooClient(
hosts,
handler=PatroniSequentialThreadingHandler(config['retry_timeout']),
timeout=config['ttl'],
connection_retry=KazooRetry(max_delay=1,
max_tries=-1,
sleep_func=time.sleep),
command_retry=KazooRetry(deadline=config['retry_timeout'],
max_delay=1,
max_tries=-1,
sleep_func=time.sleep),
**kwargs)
self._client.add_listener(self.session_listener)
self._fetch_cluster = True
self._fetch_optime = True
self._orig_kazoo_connect = self._client._connection._connect
self._client._connection._connect = self._kazoo_connect
self._client.start()
def _kazoo_connect(self, *args):
"""Kazoo is using Ping's to determine health of connection to zookeeper. If there is no
response on Ping after Ping interval (1/2 from read_timeout) it will consider current
connection dead and try to connect to another node. Without this "magic" it was taking
up to 2/3 from session timeout (ttl) to figure out that connection was dead and we had
only small time for reconnect and retry.
This method is needed to return different value of read_timeout, which is not calculated
from negotiated session timeout but from value of `loop_wait`. And it is 2 sec smaller
than loop_wait, because we can spend up to 2 seconds when calling `touch_member()` and
`write_leader_optime()` methods, which also may hang..."""
ret = self._orig_kazoo_connect(*args)
return max(self.loop_wait - 2, 2) * 1000, ret[1]
def session_listener(self, state):
if state in [KazooState.SUSPENDED, KazooState.LOST]:
self.cluster_watcher(None)
def optime_watcher(self, event):
self._fetch_optime = True
self.event.set()
def cluster_watcher(self, event):
self._fetch_cluster = True
self.optime_watcher(event)
def reload_config(self, config):
self.set_retry_timeout(config['retry_timeout'])
loop_wait = config['loop_wait']
loop_wait_changed = self._loop_wait != loop_wait
self._loop_wait = loop_wait
self._client.handler.set_connect_timeout(loop_wait)
# We need to reestablish connection to zookeeper if we want to change
# read_timeout (and Ping interval respectively), because read_timeout
# is calculated in `_kazoo_connect` method. If we are changing ttl at
# the same time, set_ttl method will reestablish connection and return
# `!True`, otherwise we will close existing connection and let kazoo
# open the new one.
if not self.set_ttl(int(config['ttl'] * 1000)) and loop_wait_changed:
self._client._connection._socket.close()
def set_ttl(self, ttl):
"""It is not possible to change ttl (session_timeout) in zookeeper without
destroying old session and creating the new one. This method returns `!True`
if session_timeout has been changed (`restart()` has been called)."""
if self._client._session_timeout != ttl:
self._client._session_timeout = ttl
self._client.restart()
return True
@property
def ttl(self):
return self._client._session_timeout
def set_retry_timeout(self, retry_timeout):
retry = self._client.retry if isinstance(
self._client.retry, KazooRetry) else self._client._retry
retry.deadline = retry_timeout
def get_node(self, key, watch=None):
try:
ret = self._client.get(key, watch)
return (ret[0].decode('utf-8'), ret[1])
except NoNodeError:
return None
def get_leader_optime(self, leader):
watch = self.optime_watcher if not leader or leader.name != self._name else None
optime = self.get_node(self.leader_optime_path, watch)
self._fetch_optime = False
return optime and int(optime[0]) or 0
@staticmethod
def member(name, value, znode):
return Member.from_node(znode.version, name, znode.ephemeralOwner,
value)
def get_children(self, key, watch=None):
try:
return self._client.get_children(key, watch)
except NoNodeError:
return []
def load_members(self, sync_standby):
members = []
for member in self.get_children(self.members_path,
self.cluster_watcher):
watch = member in sync_standby and self.cluster_watcher or None
data = self.get_node(self.members_path + member, watch)
if data is not None:
members.append(self.member(member, *data))
return members
def _inner_load_cluster(self):
self._fetch_cluster = False
self.event.clear()
nodes = set(
self.get_children(self.client_path(''), self.cluster_watcher))
if not nodes:
self._fetch_cluster = True
# get initialize flag
initialize = (self.get_node(self.initialize_path)
or [None])[0] if self._INITIALIZE in nodes else None
# get global dynamic configuration
config = self.get_node(
self.config_path,
watch=self.cluster_watcher) if self._CONFIG in nodes else None
config = config and ClusterConfig.from_node(config[1].version,
config[0], config[1].mzxid)
# get timeline history
history = self.get_node(
self.history_path,
watch=self.cluster_watcher) if self._HISTORY in nodes else None
history = history and TimelineHistory.from_node(
history[1].mzxid, history[0])
# get synchronization state
sync = self.get_node(
self.sync_path,
watch=self.cluster_watcher) if self._SYNC in nodes else None
sync = SyncState.from_node(sync and sync[1].version, sync and sync[0])
# get list of members
sync_standby = sync.leader == self._name and sync.members or []
members = self.load_members(
sync_standby) if self._MEMBERS[:-1] in nodes else []
# get leader
leader = self.get_node(
self.leader_path) if self._LEADER in nodes else None
if leader:
client_id = self._client.client_id
if not self._ctl and leader[0] == self._name and client_id is not None \
and client_id[0] != leader[1].ephemeralOwner:
logger.info(
'I am leader but not owner of the session. Removing leader node'
)
self._client.delete(self.leader_path)
leader = None
if leader:
member = Member(-1, leader[0], None, {})
member = ([m for m in members if m.name == leader[0]]
or [member])[0]
leader = Leader(leader[1].version, leader[1].ephemeralOwner,
member)
self._fetch_cluster = member.index == -1
# get last leader operation
last_leader_operation = self._OPTIME in nodes and self.get_leader_optime(
leader)
# failover key
failover = self.get_node(
self.failover_path,
watch=self.cluster_watcher) if self._FAILOVER in nodes else None
failover = failover and Failover.from_node(failover[1].version,
failover[0])
return Cluster(initialize, config, leader, last_leader_operation,
members, failover, sync, history)
def _load_cluster(self):
cluster = self.cluster
if self._fetch_cluster or cluster is None:
try:
cluster = self._client.retry(self._inner_load_cluster)
except Exception:
logger.exception('get_cluster')
self.cluster_watcher(None)
raise ZooKeeperError('ZooKeeper in not responding properly')
# Optime ZNode was updated or doesn't exist and we are not leader
elif (self._fetch_optime and not self._fetch_cluster or not cluster.last_leader_operation) and\
not (cluster.leader and cluster.leader.name == self._name):
try:
optime = self.get_leader_optime(cluster.leader)
cluster = Cluster(cluster.initialize, cluster.config,
cluster.leader, optime, cluster.members,
cluster.failover, cluster.sync,
cluster.history)
except Exception:
pass
return cluster
def _bypass_caches(self):
self._fetch_cluster = True
def _create(self, path, value, retry=False, ephemeral=False):
try:
if retry:
self._client.retry(self._client.create,
path,
value,
makepath=True,
ephemeral=ephemeral)
else:
self._client.create_async(path,
value,
makepath=True,
ephemeral=ephemeral).get(timeout=1)
return True
except Exception:
logger.exception('Failed to create %s', path)
return False
def attempt_to_acquire_leader(self, permanent=False):
ret = self._create(self.leader_path,
self._name.encode('utf-8'),
retry=True,
ephemeral=not permanent)
if not ret:
logger.info('Could not take out TTL lock')
return ret
def _set_or_create(self,
key,
value,
index=None,
retry=False,
do_not_create_empty=False):
value = value.encode('utf-8')
try:
if retry:
self._client.retry(self._client.set,
key,
value,
version=index or -1)
else:
self._client.set_async(key, value, version=index
or -1).get(timeout=1)
return True
except NoNodeError:
if do_not_create_empty and not value:
return True
elif index is None:
return self._create(key, value, retry)
else:
return False
except Exception:
logger.exception('Failed to update %s', key)
return False
def set_failover_value(self, value, index=None):
return self._set_or_create(self.failover_path, value, index)
def set_config_value(self, value, index=None):
return self._set_or_create(self.config_path, value, index, retry=True)
def initialize(self, create_new=True, sysid=""):
sysid = sysid.encode('utf-8')
return self._create(self.initialize_path, sysid, retry=True) if create_new \
else self._client.retry(self._client.set, self.initialize_path, sysid)
def touch_member(self, data, permanent=False):
cluster = self.cluster
member = cluster and cluster.get_member(self._name,
fallback_to_leader=False)
encoded_data = json.dumps(data, separators=(',', ':')).encode('utf-8')
if member and (self._client.client_id is not None
and member.session != self._client.client_id[0] or
not (deep_compare(member.data.get('tags', {}),
data.get('tags', {})) and
member.data.get('version') == data.get('version')
and member.data.get('checkpoint_after_promote')
== data.get('checkpoint_after_promote'))):
try:
self._client.delete_async(self.member_path).get(timeout=1)
except NoNodeError:
pass
except Exception:
return False
member = None
if member:
if deep_compare(data, member.data):
return True
else:
try:
self._client.create_async(
self.member_path,
encoded_data,
makepath=True,
ephemeral=not permanent).get(timeout=1)
return True
except Exception as e:
if not isinstance(e, NodeExistsError):
logger.exception('touch_member')
return False
try:
self._client.set_async(self.member_path,
encoded_data).get(timeout=1)
return True
except Exception:
logger.exception('touch_member')
return False
def take_leader(self):
return self.attempt_to_acquire_leader()
def _write_leader_optime(self, last_operation):
return self._set_or_create(self.leader_optime_path, last_operation)
def _update_leader(self):
return True
def _delete_leader(self):
self._client.restart()
return True
def _cancel_initialization(self):
node = self.get_node(self.initialize_path)
if node:
self._client.delete(self.initialize_path, version=node[1].version)
def cancel_initialization(self):
try:
self._client.retry(self._cancel_initialization)
except Exception:
logger.exception("Unable to delete initialize key")
def delete_cluster(self):
try:
return self._client.retry(self._client.delete,
self.client_path(''),
recursive=True)
except NoNodeError:
return True
def set_history_value(self, value):
return self._set_or_create(self.history_path, value)
def set_sync_state_value(self, value, index=None):
return self._set_or_create(self.sync_path,
value,
index,
retry=True,
do_not_create_empty=True)
def delete_sync_state(self, index=None):
return self.set_sync_state_value("{}", index)
def watch(self, leader_index, timeout):
if super(ZooKeeper, self).watch(leader_index,
timeout) and not self._fetch_optime:
self._fetch_cluster = True
return self._fetch_cluster
class ZookeeperServiceRegistry(BaseServiceRegistry):
def __init__(self, hosts=DEFAULT_HOSTS, chroot=DEFAULT_CHROOT):
super(ZookeeperServiceRegistry, self).__init__()
self.chroot = chroot
self.client = KazooClient(
hosts=hosts,
handler=SequentialGeventHandler(),
)
self.client.add_listener(self.on_kazoo_state_change)
self.start_count = 0
@classmethod
def from_config(cls, config, **kwargs):
return cls(
hosts=config.get('hosts', DEFAULT_HOSTS),
chroot=config.get('chroot', DEFAULT_CHROOT),
**kwargs
)
def on_start(self, timeout=10):
self.start_count += 1
if self.start_count > 1:
return
started = self.client.start_async()
started.wait(timeout=timeout)
if not self.client.connected:
raise RuntimeError('could not connect to zookeeper')
logger.debug('connected to zookeeper (version=%s)', '.'.join(map(str, self.client.server_version())))
def on_stop(self):
self.start_count -= 1
if self.start_count != 0:
return
self.client.stop()
def on_kazoo_state_change(self, state):
logger.info('kazoo connection state changed to %s', state)
def on_service_type_watch(self, service, event):
try:
if event.type == EventType.CHILD:
# FIXME: figure out proper retry strategy
self.client.retry(self.lookup, service.container, service)
except Exception:
logger.exception('error in service type watcher')
def on_service_watch(self, service, event):
try:
prefix, service_type, identity = event.path.rsplit('/', 2)
if event.type == EventType.DELETED:
service.remove(identity)
except Exception:
logger.exception('error in service watcher')
def _get_service_znode(self, service, service_type, identity):
path = self._get_zk_path(service_type, identity)
result = self.client.get_async(
path, watch=functools.partial(self.on_service_watch, service))
value, znode = result.get()
items = six.iteritems(json.loads(value.decode('utf-8')))
return {str(k): str(v) for k, v in items}
def discover(self, container):
result = self.client.get_children_async(
path='%s/services' % self.chroot,
)
return list(result.get())
def lookup(self, container, service, watch=True, timeout=1):
def child_watch(event):
print(event)
service_type = service.service_type
result = self.client.get_children_async(
path='%s/services/%s' % (self.chroot, service_type),
watch=functools.partial(self.on_service_type_watch, service),
)
try:
names = result.get(timeout=timeout)
except NoNodeError:
raise LookupFailure(None, "failed to resolve %s" % service.service_type)
logger.info("lookup %s %r", service_type, names)
identities = set(service.identities())
for name in names:
kwargs = self._get_service_znode(service, service_type, name)
identity = kwargs.pop('identity')
service.update(identity, **kwargs)
try:
identities.remove(identity)
except KeyError:
pass
for identity in identities:
service.remove(identity)
return service
def _get_zk_path(self, service_type, identity):
return '%s/services/%s/%s' % (self.chroot, service_type, identity)
def register(self, container, service_type, timeout=1):
path = self._get_zk_path(service_type, container.identity)
value = json.dumps({
'endpoint': container.endpoint,
'identity': container.identity,
'log_endpoint': container.log_endpoint,
})
result = self.client.create_async(
path,
value.encode('utf-8'),
ephemeral=True, makepath=True)
# FIXME: result.set_exception(RegistrationFailure())
result.get(timeout=timeout)
def unregister(self, container, service_type, timeout=1):
path = self._get_zk_path(service_type, container.identity)
result = self.client.delete_async(path)
result.set_exception(RegistrationFailure())
result.get(timeout=timeout)
MASTER_COUNT += 1
SLAVE_COUNT += 1
WORKER_COUNT += 1
container = client.containers.run(
"workers:latest",
detach=True,
name="worker_container" + str(WORKER_COUNT),
network="orch-network",
command=["sh", "-c", "service mongodb start; python3 worker.py 1"])
pid = p_client.inspect_container(container.name)['State']['Pid']
MASTER_LIST.append(container)
message = ("running " + str(pid)).encode()
zk.create("/worker/master", message, makepath=True)
zk.create("/worker/master/" + str(pid), b"running")
WORKER_COUNT += 1
container = client.containers.run(
"workers:latest",
detach=True,
name="worker_container" + str(WORKER_COUNT),
network="orch-network",
command=["sh", "-c", "service mongodb start; python3 worker.py 0"])
pid = p_client.inspect_container(container.name)['State']['Pid']
SLAVE_LIST.append(container)
message = ("running " + str(pid)).encode()
zk.create_async("/worker/slave", message, makepath=True)
zk.create_async("/worker/slave/" + str(pid), b"running")
app.run(host='0.0.0.0', debug=False)
class Zookeeper:
def __init__(self, hosts):
self.zk = KazooClient(hosts=hosts,
handler=SequentialGeventHandler(),
logger=logger)
# returns immediately
event = self.zk.start_async()
# Wait for 30 seconds and see if we're connected
event.wait(timeout=30)
try:
if not self.zk.connected:
# Not connected, stop trying to connect
self.zk.stop()
except (ConnectionLossException, NoAuthException) as error:
raise error
except Exception as error:
raise error
@coroutine
def get_children(self, node):
try:
children = self.zk.get_children_async(node)
raise Return(children.get())
except Exception as error:
raise error
@coroutine
def get_node(self, node):
try:
data = self.zk.get_async(node)
raise Return(data.get())
except Exception as error:
raise error
@coroutine
def check_path_exist(self, path):
try:
result = self.zk.exists(path)
if result:
raise Return(True)
else:
raise Return(False)
except Exception as error:
raise error
@coroutine
def create_path(self, path):
try:
result = self.zk.ensure_path_async(path)
raise Return(result.get())
except Exception as error:
raise error
@coroutine
def create_node(self, path, value):
try:
result = self.zk.create_async(path=path,
value=value,
acl=None,
ephemeral=True)
raise Return(result.get())
except Exception as error:
raise error
@coroutine
def update_node(self, path, value, version=-1):
try:
result = self.zk.set_async(path, value, version)
raise Return(result.get())
except Exception as error:
raise error
@coroutine
def update_node(self, path, value, version=-1):
try:
result = self.zk.set_async(path, value, version)
raise Return(result.get())
except Exception as error:
raise error
@coroutine
def del_node(self, node):
try:
node_info = self.zk.delete_async(node)
raise Return(node_info.get())
except Exception as error:
raise error
def close(self):
self.zk.stop()
class Zooconf:
connection = None
webService = None
applicationService = None
storageServicesList = None
authenticationServiceList = None
status = None
def initialize(self):
self.__zooConnect()
self.__publishService()
self.__initAuthenticationServiceWatches()
self.__initStorageServiceWatches()
self.heartbeat()
def __zooConnect(self):
print("Connecting to ZooKeeper")
self.connection = KazooClient(hosts=settings.ZOOKEEPER_HOST)
self.connection.start()
digest_auth = "%s:%s" % (settings.ZOOKEEPER_USER,
settings.ZOOKEEPER_PASSWORD)
self.connection.add_auth("digest", digest_auth)
def __publishService(self):
acl = make_digest_acl(settings.ZOOKEEPER_USER,
settings.ZOOKEEPER_PASSWORD,
all=True)
dataJsonDict = {
'SERVER_HOSTNAME': settings.SERVER_HOSTNAME,
'SERVER_PORT': settings.SERVER_PORT,
'SHARED_KEY_BASE_64': settings.SHARED_KEY_BASE_64,
'CHILDREN': []
}
if settings.ZOOKEEPER_PATH_TO_NODE == 'Auth/':
dataJsonDict['AUTH_SYSTEM'] = settings.AUTH_SYSTEM
if self.connection.exists(path=settings.ZOOKEEPER_ROOT +
settings.ZOOKEEPER_PATH_TO_NODE +
settings.ZOOKEEPER_NODE_ID):
self.connection.set(
path=settings.ZOOKEEPER_ROOT +
settings.ZOOKEEPER_PATH_TO_NODE + settings.ZOOKEEPER_NODE_ID,
value=json.JSONEncoder().encode(dataJsonDict).encode())
else:
self.connection.create_async(
path=settings.ZOOKEEPER_ROOT +
settings.ZOOKEEPER_PATH_TO_NODE + settings.ZOOKEEPER_NODE_ID,
value=json.JSONEncoder().encode(dataJsonDict).encode(),
ephemeral=settings.ZOOKEEPER_NODE_EPHIMERAL)
if settings.ZOOKEEPER_PATH_TO_NODE != '':
data, stat = self.connection.get(settings.ZOOKEEPER_ROOT +
settings.ZOOKEEPER_PATH_TO_NODE)
dataJsonDict = json.loads(data.decode("utf-8"))
if settings.ZOOKEEPER_NODE_ID not in dataJsonDict['CHILDREN']:
dataJsonDict['CHILDREN'].append(settings.ZOOKEEPER_NODE_ID)
self.connection.set(
path=settings.ZOOKEEPER_ROOT + settings.ZOOKEEPER_PATH_TO_NODE,
value=json.JSONEncoder().encode(dataJsonDict).encode())
def __initStorageServiceWatches(self):
self.storageServicesList = []
@self.connection.ChildrenWatch(settings.ZOOKEEPER_ROOT +
"StorageServices")
def watch_children(children):
print(self.readTree())
if settings.ZOOKEEPER_PATH_TO_NODE == "StorageServices/":
node = {
'SERVER_HOSTNAME': settings.SERVER_HOSTNAME,
'SERVER_PORT': settings.SERVER_PORT,
'CHILDREN': []
}
if node not in self.storageServicesList:
self.__publishService()
def __initAuthenticationServiceWatches(self):
self.authenticationServiceList = []
@self.connection.ChildrenWatch(settings.ZOOKEEPER_ROOT + "Auth")
def watch_children(children):
print(self.readTree())
if settings.ZOOKEEPER_PATH_TO_NODE == "Auth/":
node = {
'SERVER_HOSTNAME': settings.SERVER_HOSTNAME,
'SERVER_PORT': settings.SERVER_PORT,
'SHARED_KEY_BASE_64': settings.SHARED_KEY_BASE_64,
'CHILDREN': [],
'AUTH_SYSTEM': settings.AUTH_SYSTEM
}
if node not in self.authenticationServiceList:
self.__publishService()
def getAvailableFs(self):
return self.storageServicesList
def getZooConnection(self):
return self.connection
def readTree(self):
result = "{"
rootChildren = self.connection.get_children(settings.ZOOKEEPER_ROOT)
for child in rootChildren:
data, stat = self.connection.get(settings.ZOOKEEPER_ROOT + child)
result += '"' + child + '": ' + data.decode("utf-8") + ","
try:
grandchildren = self.connection.get_children(
settings.ZOOKEEPER_ROOT + child)
for grandchild in grandchildren:
data, stat = self.connection.get(settings.ZOOKEEPER_ROOT +
child + '/' + grandchild)
result += '"' + grandchild + '": ' + data.decode(
"utf-8") + ","
data, stat = self.connection.get(settings.ZOOKEEPER_ROOT +
child + '/')
dataJsonDict = json.loads(data.decode("utf-8"))
dataJsonDict['CHILDREN'] = grandchildren
self.connection.set(
path=settings.ZOOKEEPER_ROOT + child,
value=json.JSONEncoder().encode(dataJsonDict).encode())
except Exception:
pass
result = result[:-1] + '}'
self.status = json.loads(result)
self.initZkTree()
return self.status
def getStatus(self):
try:
return self.readTree()
except Exception:
self.__zooConnect()
return self.getStatus()
def getNodeData(self, node):
try:
return self.status[node]
except Exception:
return {}
def initZkTree(self):
serviceData = self.getNodeData('WebService')
self.webService = serviceData['SERVER_HOSTNAME'] + ':' + serviceData[
'SERVER_PORT'] + '/'
serviceData = self.getNodeData('ApplicationService')
self.applicationService = serviceData[
'SERVER_HOSTNAME'] + ':' + serviceData['SERVER_PORT'] + '/'
serviceData = self.getNodeData('Auth')
self.authenticationServiceList = []
for authService in serviceData['CHILDREN']:
authServiceData = self.getNodeData(authService)
if authServiceData != {}:
baseUrl = authServiceData['SERVER_HOSTNAME'] + (
(':' + authServiceData['SERVER_PORT'])
if authServiceData['SERVER_PORT'] != '' else '') + '/'
loginUrl = baseUrl + 'login?system=' + authServiceData[
'AUTH_SYSTEM'] + '&callback='
registerUrl = baseUrl + 'register?system=' + authServiceData[
'AUTH_SYSTEM'] + '&callback='
authServiceDict = {
'name': authService,
'url': baseUrl,
'loginUrl': loginUrl,
'registerUrl': registerUrl,
'sharedKey': authServiceData['SHARED_KEY_BASE_64'],
'system': authServiceData['AUTH_SYSTEM']
}
self.authenticationServiceList.append(authServiceDict)
serviceData = self.getNodeData('StorageServices')
self.storageServicesList = []
for storageService in serviceData['CHILDREN']:
storageServiceData = self.getNodeData(storageService)
if storageServiceData != {}:
storageServiceDict = {
'name':
storageService,
'url':
storageServiceData['SERVER_HOSTNAME'] + ':' +
storageServiceData['SERVER_PORT'] + '/'
}
self.storageServicesList.append(storageServiceDict)
def heartbeat(self):
threading.Timer(300.0, self.heartbeat).start()
print("Heartbeat")
print(str(self.getStatus()))
class ZookeeperLeaderElector:
zk = None
default_node_path = None
zookeeper_connection_url = None
config_file_dir = None
def __init__(self):
# Default zookeeper node path to CRUD
self.default_node_path = "/cluster"
# Locate config file
self._get_config_file_addr()
# Get the subset of properties relevant to zookeeper
self._init_config_props()
# Init zookeeper client instance
self.zk = KazooClient(self.zookeeper_connection_url)
# Locate config file
def _get_config_file_addr(self):
current_dir = path.dirname(path.realpath(__file__))
parent_dir = path.dirname(current_dir)
self.config_file_dir = path.join(path.dirname(parent_dir), 'config',
'zookeeper.config')
# Get the subset of properties relevant to zookeeper
def _init_config_props(self):
server_props = ConfigParser()
server_props.read(self.config_file_dir)
self.zookeeper_connection_url = server_props["connect"]["url"]
def startup(self, server_id, host_ip, broker_server_start):
# Connect to zookeeper server
self.zk.start()
try:
# Ensure a path, create if necessary
self.zk.ensure_path(self.default_node_path)
# Create a node with data
node = "node" + server_id
node_path = self.default_node_path + '/' + node
self.zk.create_async(node_path,
bytes(host_ip, 'utf-8'),
ephemeral=True)
# Elect for leadership
print("[SETUP/ZK] Elect for leadership ...")
election = self.zk.Election(self.default_node_path, node)
election.run(broker_server_start)
# Exit
except KeyboardInterrupt:
self.exit()
# Alwasys stop the zk instance and disconnect
finally:
self.zk.stop()
self.zk.close()
def exit(self):
print("[EXIT] Disconnect from zookeeper server.")
raise KeyboardInterrupt
def ready(self):
if not self.config_file_dir:
sys.exit("[ERR] Doesn't find the config file.")
elif not self.zookeeper_connection_url:
sys.exit("[ERR] Zookeeper server url is EMPTY.")
elif not self.zk:
sys.exit("[ERR] Zookeeper instance instantiation FAILED.")
return True
class ZooHandler(object):
def __init__(self):
self.zookeeper_client = None
if not settings.ZOOKEEPER_SETTING['enable']:
logging.info('zookeeper disabled')
return
self.zoo_hosts = settings.ZOOKEEPER_SETTING['server_address']
logging.info('start zookeeper client, zoo hosts: %s' % self.zoo_hosts)
self.base_dir = settings.ZOOKEEPER_SETTING['base_dir']
self.zookeeper_client = KazooClient(hosts=self.zoo_hosts)
self.zookeeper_client.add_listener(self.state_listener)
self.zookeeper_client.start_async()
def state_listener(self, state):
# session was lost
if state == KazooState.LOST:
logging.error('zookeeper lost!')
# disconnected from Zookeeper
elif state == KazooState.SUSPENDED:
logging.error('zookeeper disconnected!')
# connected/reconnected to Zookeeper
elif state == KazooState.CONNECTED:
self.register_node()
logging.warn('zookeeper reconnected! try to register')
else:
logging.error('unexpected zookeeper state!!!')
logging.critical('unexpected zookeeper state!!!')
def register_node(self):
if not self.zookeeper_client or not self.zookeeper_client.connected:
logging.error('zoo not connected, register cancel')
return
path = ZooHandler.get_register_path()
try:
# 尝试注册节点
def try_to_create_node(result):
logging.info('zoo try_to_create_noe called')
try:
# None表示节点不存在
if result.value is None:
self.zookeeper_client.create_async(path,
makepath=True,
ephemeral=True)
elif result.exception:
logging.fatal(
'critical error when try to check node when reconnected, %s',
result.exception)
else:
logging.warn(
'node already exists when reconnect and try to register'
)
except BaseException as e:
logging.exception('critical error, %s', e.message)
# 监控节点变化
def node_watcher(watch_event):
logging.info('zoo node_watcher called')
try:
if EventType.DELETED == watch_event.type:
logging.warn('zoo nodes deleted, try recreate')
self.zookeeper_client.create_async(path,
makepath=True,
ephemeral=True)
if EventType.CHANGED == watch_event.type:
logging.warn('zoo nodes changed,do nothing')
if EventType.CHILD == watch_event.type:
logging.warn('zoo nodes childed,do nothing')
if EventType.CREATED == watch_event.type:
logging.info('zoo nodes success created')
if EventType.NONE == watch_event.type:
logging.error('zoo nodes status return None')
finally:
self.zookeeper_client.exists_async(path,
watch=node_watcher)
future = self.zookeeper_client.exists_async(path,
watch=node_watcher)
future.rawlink(try_to_create_node)
except ZookeeperError as e:
logging.exception('zookeeper exception when register node: %s' %
e.message)
except BaseException as e:
logging.exception('critical error!')
# 1. remove nodes,stop client
def stop(self):
logging.info('stopping zookeeper client')
if self.zookeeper_client:
self.zookeeper_client.remove_listener(self.state_listener)
self.zookeeper_client.stop()
logging.info('zookeeper stopped')
@staticmethod
def get_register_path():
base_dir = settings.ZOOKEEPER_SETTING['base_dir']
if base_dir[-1] == '/':
base_dir = base_dir[0:-1]
register_name = "%s/%s:%s:%s" % (
base_dir, settings.ZOOKEEPER_SETTING['local_name'],
settings.ZOOKEEPER_SETTING['local_ip'],
settings.HTTP_SERVER_SETTING['port'])
return register_name
timeout=1,
handler=SequentialGeventHandler())
event = zk.start_async()
event.wait(timeout=1) # wait()方法等待start_async()返回的事件对象
if not zk.connected: # 由于可能永远连接失败，因此判断连接状态，做异常情况处理
zk.stop()
raise Exception("Unable to connect")
def my_callback(async_obj):
try:
print '-------------------------'
children = async_obj.get()
do_something(children)
except (ConnectionLossException, NoAuthException):
sys.exit(1)
zk.create_async("/lws/test/1", b"test")
data = zk.get_async("/lws/test/1")
print data
async_obj = zk.get_children_async("/lws/test/1")
# print async_obj
async_obj.rawlink(my_callback)
# data = zk.exists_async("/lws/test/1")
# print data
class ZookeeperNonBrokerManager:
zk = None
default_node_path = None
config_file_dir = None
zookeeper_connection_url = None
zookeeper_connection_timeout = None
# subscriber only
publisher_server_default_port = None
publisher_server_url_lst = list()
def __init__(self, role):
print("[SETUP/ZK] Communicate NOT through broker")
# Default zookeeper node path to CRUD
self.default_node_path = "/publishers"
# Get config file
self._get_config_file_addr(role)
# Get the subset of properties relevant to zookeeper
self._init_config_props(role)
# Init zookeeper client instance
self.zk = KazooClient(self.zookeeper_connection_url)
# Locate config file
def _get_config_file_addr(self, role):
current_dir = path.dirname(path.realpath(__file__))
parent_dir = path.dirname(path.dirname(current_dir))
filename = role + ".config"
self.config_file_dir = path.join(path.dirname(parent_dir), "config",
filename)
# Get the subset of properties relevant to zookeeper
def _init_config_props(self, role):
props = ConfigParser()
props.read(self.config_file_dir)
self.zookeeper_connection_url = props["service_discovery"][
"connection.url"]
self.zookeeper_connection_timeout = int(
props["service_discovery"]["connection.timeout.s"])
if role == "publisher":
pass
elif role == "subscriber":
self.publisher_server_default_port = props["publisher_server"][
"port"]
else:
sys.exit("[ERR] Role not exists")
def publisher_connect(self, role, node_id, host_ip):
# Call prevention
if role != "publisher":
sys.exit("[ERR] This method can only be called by publisher")
# Connect to zookeeper server
if not self.zk.connected:
self.zk.start(timeout=self.zookeeper_connection_timeout)
try:
# Ensure a path, create if necessary
self.zk.ensure_path(self.default_node_path)
# Create a node with data
node = "node" + node_id
node_path = self.default_node_path + '/' + node
self.zk.create_async(node_path,
bytes(host_ip, "utf-8"),
ephemeral=True)
# Exit
except KeyboardInterrupt:
self.exit()
def subscriber_connect(self, role, socks_connect, socks_disconnect):
# Call prevention
if role != "subscriber":
sys.exit("[ERR] This method can only be called by subscriber")
# Connect to zookeeper server
if not self.zk.connected:
self.zk.start(timeout=self.zookeeper_connection_timeout)
try:
# Ensure a path, create if necessary
self.zk.ensure_path(self.default_node_path)
# Watch on children
self.watch(socks_connect, socks_disconnect)
# Exit
except KeyboardInterrupt:
self.exit()
# Watch on the leader node, find new leader if the current leader suicides
def watch(self, socks_connect, socks_disconnect):
@self.zk.ChildrenWatch(self.default_node_path)
def my_func(updated_publisher_server_url_lst):
if sorted(updated_publisher_server_url_lst) != sorted(
self.publisher_server_url_lst):
# Find publishers to disconeect
deleted_publisher_servers = [
node for node in self.publisher_server_url_lst
if node not in updated_publisher_server_url_lst
]
# Find newly join publishers to connect
new_publisher_servers = [
node for node in updated_publisher_server_url_lst
if node not in self.publisher_server_url_lst
]
# Update local publisher_server_url_lst
self.publisher_server_url_lst = updated_publisher_server_url_lst
if deleted_publisher_servers:
print("[SETUP/ZK] Disconnected from " +
str(deleted_publisher_servers))
for node in deleted_publisher_servers:
node_path = self.default_node_path + '/' + node
data, _ = self.zk.get(node_path)
host_ip = "tcp://{0}:{1}".format(
data.decode("utf-8"),
self.publisher_server_default_port)
print("disconnect from: " + host_ip)
socks_disconnect(host_ip)
if new_publisher_servers:
print("[SETUP/ZK] Connected to " +
str(new_publisher_servers))
for node in new_publisher_servers:
node_path = self.default_node_path + '/' + node
data, _ = self.zk.get(node_path)
host_ip = "tcp://{0}:{1}".format(
data.decode("utf-8"),
self.publisher_server_default_port)
print("connect to: " + host_ip)
socks_connect(host_ip)
self.watch(socks_connect, socks_disconnect)
else:
print("[SETUP/ZK] Start watching on leader node")
def ready(self):
if not self.config_file_dir:
sys.exit("[ERR] Doesn't find the config file.")
elif not self.zookeeper_connection_url:
sys.exit("[ERR] Zookeeper server url is EMPTY.")
elif not self.zk:
sys.exit("[ERR] Zookeeper instance instantiation FAILED.")
return True
def exit(self):
print("[EXIT] Disconnect from zookeeper server.")
# Alwasys stop the zk instance and disconnect
self.zk.stop()
self.zk.close()
raise KeyboardInterrupt
class Coordinator(object):
def __init__(self, zkhosts, root=NODE_HQ_ROOT, alivenode="alive", readonly=False, role=None):
"""zkhosts: a string or a list. list will be ','.join-ed into a string.
root: root node path (any parents must exist, if any)
"""
self.LOGGER = logging.getLogger("hq.zkcoord")
if not isinstance(zkhosts, basestring):
zkhosts = ",".join(zkhosts)
self.zkhosts = zkhosts
self.ROOT = root
self.alivenode = alivenode
self.readonly = readonly
self.nodename = os.uname()[1]
self.NODE_SERVERS = self.ROOT + "/servers"
self.NODE_ME = self.NODE_SERVERS + "/" + self.nodename
self.NODE_MYJOBS = self.NODE_ME + "/jobs"
self.NODE_GJOBS = self.ROOT + "/jobs"
self.__listeners = {}
self.jobs = {}
self.zh = None
self.zstate = None
self._connect()
def _connect(self):
try:
if self.zh:
self.zh.stop()
self.LOGGER.debug("connecting to %s", self.zkhosts)
self.zh = KazooClient(hosts=self.zkhosts)
self.zh.add_listener(self.__watcher)
# this will wait until connection is established.
self.zh.start()
except ZookeeperError as ex:
self.zh = None
self.zkerror = ex
def _initialize(self):
if not self.readonly:
if self.zstate is None:
self.zh.ensure_path(self.NODE_SERVERS)
self.zh.ensure_path(self.NODE_GJOBS)
self.publish_alive()
if not self.zh.exists(self.NODE_ME):
self.zh.create(self.NODE_ME, "")
if not self.zh.exists(self.NODE_MYJOBS):
self.zh.acreate(self.NODE_MYJOBS)
# setup notifications
self.zh.get_children(self.NODE_SERVERS, self.__servers_watcher)
self.zh.get_children(self.NODE_GJOBS, self.__jobs_watcher)
def __watcher(self, state):
# client level callback method. this method should not block.
if state == KazooState.LOST:
# session expiration
self.zstate = state
elif state == KazooState.SUSPENDED:
# disconnected, session is still alive
self.zstate = state
else:
# (re)connected
self.LOGGER.debug("connected")
self.zh.handler.spawn(self._initialize)
self.zstate = state
def get_status_text(self):
return self.zkerror
def create(self, path, data=""): # , perm=PERM_WORLD, flags=''):
return self.zh.create(path, data)
def acreate(self, path, data=""): # , perm=PERM_WORLD, flags=''):
return self.zh.acreate(path, data)
def exists(self, path):
return self.zh.exists(path)
def delete(self, path):
try:
return self.zh.delete(path)
except NoNodeError as ex:
pass
def get_children(self, path, watch=None):
return self.zh.get_children(path, watch=watch)
def __servers_watcher(self, zh, evtype, state, path):
"""called when HQ servers are added / dropped."""
try:
ch = self.get_children(self.NODE_SERVERS, watch=self.__servers_watcher)
self.LOGGER.info("servers added/removed:%s", str(ch))
self.fire_event("serverschanged")
except ZookeeperError as ex:
self.LOGGER.warn("zk.get_children(%r) failed", self.NODE_SERVERS, exc_info=1)
def __jobs_watcher(self, zh, evtype, state, path):
"""called when jobs are added / dropped."""
try:
self.LOGGER.info("%s children changed", self.NODE_GJOBS)
ch = self.get_children(self.NODE_GJOBS, watch=self.__jobs_watcher)
self.fire_event("jobschanged")
except ZooKeeperError as ex:
self.LOGGER.warn("get_children(%r) failed", self.NODE_GJOBS, exc_info=1)
def publish_alive(self):
node_alive = self.NODE_ME + "/" + self.alivenode
self.zh.create_async(node_alive, ephemeral=True)
def publish_job(self, job):
"""job: hq.CrawlJob"""
ju = self.jobs.get(job)
# update 10 minutes interval
if ju is None or ju < time.time() - 10 * 60:
NODE_MYJOB = self.NODE_MYJOBS + "/" + job.jobname
def set_complete(a):
# print >>sys.stderr, "aset completed: %s" % str(args)
if a.exception == NoNodeError:
# node does not exist yet - create anew
self.zh.create_async(NODE_MYJOB, "")
try:
a = self.zh.set_async(NODE_MYJOB, "")
a.rawlink(set_complete)
except:
self.LOGGER.warn("aset failed", exc_info=1)
pass
node2 = self.NODE_GJOBS + "/" + job.jobname
self.zh.create_async(node2)
self.zh.create_async("{}/{}/{}".format(self.NODE_GJOBS, job.jobname, self.nodename), ephemeral=True)
self.jobs[job] = time.time()
def publish_client(self, job, client):
pass
def get_servers(self):
return self.zh.get_children(self.NODE_SERVERS)
def get_server_job(self, server, job):
p = self.NODE_SERVERS + "/" + server + "/jobs/" + job
j = dict()
try:
nodeval = self.zh.get(p)
attr = nodeval[1]
j["ts"] = attr.mtime / 1000.0
except NoNodeError as ex:
j["ts"] = 0
return j
def get_status_of(self, server=None, jobs=None):
if self.zh is None:
return None
server = server or self.nodename
status = dict(name=server)
try:
node = self.zh.get(self.NODE_SERVERS + "/" + server + "/alive")
status["alive"] = node[1]
except NoNodeError as ex:
status["alive"] = False
jobspath = self.NODE_SERVERS + "/" + server + "/jobs"
if jobs is None:
try:
jobs = self.get_children(jobspath)
except NoNodeError:
jobs = []
elif isinstance(jobs, basestring):
jobs = [jobs]
status["jobs"] = []
for j in jobs:
jobj = self.get_server_job(server, j)
jobj["name"] = j
status["jobs"].append(jobj)
return status
def get_servers_status(self):
return [self.get_status_of(server) for server in self.get_servers()]
def get_job_servers(self, jobname):
"""return a map of integer identifier to server name, which
is configured for the job jobname.
"""
p = self.NODE_GJOBS + "/" + jobname + "/servers"
try:
svids = self.get_children(p)
except NoNodeError as ex:
return {}
servers = {}
for name in svids:
try:
svid = int(name)
except:
continue
nodevals = self.zh.get(p + "/" + str(svid))
if nodevals[0]:
servers[svid] = nodevals[0]
return servers
def get_job_servers2(self, jobname):
"""returns servers for job "jobname", including those active
but not registered at jobs/JOBNAME/servers. elements of returned
list are dict with svid and name keys. svid key only exists for
those registered for the "jobname".
"""
servers = [dict(svid=svid, name=name) for svid, name in self.get_job_servers(jobname).items()]
regservers = set(s["name"] for s in servers)
try:
p = self.NODE_SERVERS
ss = self.get_children(self.NODE_SERVERS)
for s in ss:
if s in regservers:
continue
if not self.is_server_alive(s):
continue
p = self.NODE_SERVERS + "/" + s + "/jobs/" + jobname
if self.exists(p):
servers.append(dict(name=s))
except ZookeeperError as ex:
self.LOGGER.debug("zookeeper access failed", exc_info=1)
return servers
def add_job_server(self, job, server):
pass
def delete_job_server(self, job, server):
jobservers = dict((v, k) for k, v in self.get_job_servers(job).items())
if server in jobservers:
p = self.NODE_GJOBS + "/" + job + "/servers/" + jobservers[server]
self.delete(p)
p = self.NODE_SERVERS + "/" + server + "/jobs/" + job
# assumption: there's no child under the server/job node.
try:
self.delete(p)
except NoNodeError as ex:
pass
except NotEmptyException as ex:
# XXX
pass
def is_server_alive(self, server):
p = self.NODE_SERVERS + "/" + server + "/alive"
return self.exists(p)
def add_listener(self, ev, listener):
if not isinstance(ev, basestring):
raise ValueError, "ev must be a string"
ll = self.__listeners.get(ev)
if not ll:
self.__listeners[ev] = set((listener,))
else:
ll.add(listener)
def remove_listener(self, ev, listener):
if not isinstance(ev, basestring):
raise ValueError, "ev must be a string"
ll = self.__listeners.get(ev)
if ll:
try:
ll.remove(listener)
except KeyError:
pass
def fire_event(self, ev, *args):
ll = self.__listeners.get(ev)
if ll:
for listener in ll:
try:
listener(*args)
except:
self.LOGGER.warn("error running listener %r " "with ev=%r, args=%r", listener, ev, args, exc_info=1)
def shutdown(self):
if self.zh:
self.zh.stop()
self.zh = None
from kazoo.handlers.gevent import SequentialGeventHandler
logging.basicConfig()
zk = KazooClient(hosts='zoo1:2181', handler=SequentialGeventHandler())
event = zk.start_async()
event.wait(timeout=30)
if not zk.connected:
# Not connected, stop trying to connect
print("ummm")
zk.stop()
raise Exception("Unable to connect.")
def my_callback(async_obj):
try:
children = async_obj.get()
print("hi")
except (ConnectionLossException, NoAuthException):
sys.exit(1)
# Both these statements return immediately, the second sets a callback
# that will be run when get_children_async has its return value
zk.create_async("/master", b'mas')
zk.create_async("/master/node_1", b'c1')
print("gdgd")
async_obj = zk.get_children_async("/master")
async_obj.rawlink(my_callback)
zk.create_async("/master/node_2", b'c2')