1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
|
from functools import partial
import contextlib
import logging
import time
import zookeeper
from twisted.internet.defer import (
inlineCallbacks, DeferredLock, fail, returnValue, Deferred)
from client import (
ZookeeperClient, ClientEvent, NotConnectedException,
ConnectionTimeoutException)
from retry import RetryClient, is_session_error
from utils import sleep
class StopWatcher(Exception):
pass
BACKOFF_INCREMENT = 10
MAX_BACKOFF = 360
WATCH_KIND_MAP = {
"child": "get_children_and_watch",
"exists": "exists_and_watch",
"get": "get_and_watch"}
log = logging.getLogger("txzk.managed")
#log.setLevel(logging.INFO)
class Watch(object):
"""
For application driven persistent watches, where the application
is manually resetting the watch.
"""
__slots__ = ("_mgr", "_client", "_path", "_kind", "_callback")
def __init__(self, mgr, path, kind, callback):
self._mgr = mgr
self._path = path
self._kind = kind
self._callback = callback
@property
def path(self):
return self._path
@property
def kind(self):
return self._kind
@contextlib.contextmanager
def _ctx(self):
mgr = self._mgr
del self._mgr
try:
yield mgr
finally:
mgr.remove(self)
@inlineCallbacks
def reset(self):
with self._ctx():
yield self._callback(
zookeeper.SESSION_EVENT,
zookeeper.CONNECTED_STATE,
self._path)
def __call__(self, *args, **kw):
with self._ctx():
return self._callback(*args, **kw)
def __str__(self):
return "<Watcher %s %s %r>" % (self.kind, self.path, self._callback)
class WatchManager(object):
watch_class = Watch
def __init__(self):
self._watches = []
def add(self, path, watch_type, watcher):
w = self.watch_class(self, path, watch_type, watcher)
self._watches.append(w)
return w
def remove(self, w):
try:
self._watches.remove(w)
except ValueError:
pass
def iterkeys(self):
for w in self._watches:
yield (w.path, w.kind)
def clear(self):
del self._watches
self._watches = []
@inlineCallbacks
def reset(self, *ignored):
watches = self._watches
self._watches = []
for w in watches:
try:
yield w.reset()
except Exception, e:
log.error("Error reseting watch %s with session event. %s %r",
w, e, e)
continue
class SessionClient(ZookeeperClient):
"""A managed client that automatically re-establishes ephemerals and
triggers watches after reconnecting post session expiration.
This abstracts the client from session expiration handling. It does
come at a cost though.
There are two application constraints that need to be considered for usage
of the SessionClient or ManagedClient. The first is that watch callbacks
which examine the event, must be able to handle the synthetic session
event which is sent to them when the session is re-established.
The second and more problematic is that algorithms/patterns
utilizing ephemeral sequence nodes need to be rethought, as the
session client will recreate the nodes when reconnecting at their
previous paths. Some algorithms (like the std zk lock recipe) rely
on properties like the smallest valued ephemeral sequence node in
a container to identify the lock holder, with the notion that upon
session expiration a new lock/leader will be sought. Sequence
ephemeral node recreation in this context is problematic as the
node is recreated at the exact previous path. Alternative lock
strategies that do work are fairly simple at low volume, such as
owning a particular node path (ie. /locks/lock-holder) with an
ephemeral.
As a result the session client only tracks and restablishes non sequence
ephemeral nodes. For coordination around ephemeral sequence nodes it
provides for watching for the establishment of new sessions via
`subscribe_new_session`
"""
def __init__(self, servers=None, session_timeout=None,
connect_timeout=4000):
"""
"""
super(SessionClient, self).__init__(servers, session_timeout)
self._connect_timeout = connect_timeout
self._watches = WatchManager()
self._ephemerals = {}
self._session_notifications = []
self._reconnect_lock = DeferredLock()
self.set_connection_error_callback(self._cb_connection_error)
self.set_session_callback(self._cb_session_event)
self._backoff_seconds = 0
self._last_reconnect = time.time()
def subscribe_new_session(self):
d = Deferred()
self._session_notifications.append(d)
return d
@inlineCallbacks
def cb_restablish_session(self, e=None, forced=False):
"""Called on intercept of session expiration to create new session.
This will reconnect to zk, re-establish ephemerals, and
trigger watches.
"""
yield self._reconnect_lock.acquire()
log.debug(
"Connection reconnect, lock acquired handle:%d", self.handle)
try:
# If its been explicitly closed, don't re-establish.
if self.handle is None:
log.debug("No handle, client closed")
return
# Don't allow forced reconnect hurds within a session.
if forced and (
(time.time() - self._last_reconnect)
< self.session_timeout / 1000.0):
forced = False
if not forced and not self.unrecoverable:
log.debug("Client already connected, allowing retry")
return
elif self.connected or self.handle >= 0:
self.close()
self.handle = -1
# Re-establish
yield self._cb_restablish_session().addErrback(
self._cb_restablish_errback, e)
except Exception, e:
log.error("error while re-establish %r %s" % (e, e))
finally:
log.debug("Reconnect lock released %s", self)
yield self._reconnect_lock.release()
@inlineCallbacks
def _cb_restablish_session(self):
"""Re-establish a new session, and recreate ephemerals and watches.
"""
# Reconnect
while 1:
log.debug("Reconnect loop - connect timeout %d",
self._connect_timeout)
# If we have some failures, back off
if self._backoff_seconds:
log.debug("Backing off reconnect %d" % self._backoff_seconds)
yield sleep(self._backoff_seconds)
# The client was explicitly closed, abort reconnect.
if self.handle is None:
returnValue(self.handle)
try:
yield self.connect(timeout=self._connect_timeout)
log.info("Restablished connection")
self._last_reconnect = time.time()
except ConnectionTimeoutException:
log.debug("Timeout establishing connection, retrying...")
except zookeeper.ZooKeeperException, e:
log.exception("Error while connecting %r %s" % (e, e))
except Exception, e:
log.info("Reconnect unknown error, aborting: %s", e)
raise
else:
break
if self._backoff_seconds < MAX_BACKOFF:
self._backoff_seconds = min(
self._backoff_seconds + BACKOFF_INCREMENT,
MAX_BACKOFF - 1)
# Recreate ephemerals
items = self._ephemerals.items()
self._ephemerals = {}
for path, e in items:
try:
yield self.create(
path, e['data'], acls=e['acls'], flags=e['flags'])
except zookeeper.NodeExistsException:
log.error("Attempt to create ephemeral node failed %r", path)
# Signal watches
yield self._watches.reset()
# Notify new session observers
notifications = self._session_notifications
self._session_notifications = []
# all good, reset backoff
self._backoff_seconds = 0
for n in notifications:
n.callback(True)
def _cb_restablish_errback(self, err, failure):
"""If there's an error re-establishing the session log it.
"""
log.error("Error while trying to re-establish connection %s\n%s" % (
err, failure))
return failure
@inlineCallbacks
def _cb_connection_error(self, client, error):
"""Convert session expiration to a transient connection error.
Dispatches from api usage error.
"""
if not is_session_error(error):
raise error
log.debug("Connection error detected, delaying retry...")
yield sleep(1)
raise zookeeper.ConnectionLossException
# Dispatch from retry exceed session maximum
def cb_retry_expired(self, error):
log.debug("Persistent retry error, reconnecting...")
return self.cb_restablish_session(forced=True)
# Dispatch from connection events
def _cb_session_event(self, client, event):
if (event.type == zookeeper.SESSION_EVENT and
event.connection_state == zookeeper.EXPIRED_SESSION_STATE):
log.debug("Client session expired event, restablishing")
self.cb_restablish_session()
# Client connected tracker on client operations.
def _check_connected(self, d):
"""Clients are automatically reconnected."""
if self.connected:
return
if self.handle is None:
d.errback(NotConnectedException("Connection closed"))
return d
log.info("Detected dead connection, reconnecting...")
c_d = self.cb_restablish_session()
def after_connected(client):
"""Return a transient connection failure.
The retry client will automatically attempt to retry the operation.
"""
log.debug("Reconnected, returning transient error")
return fail(zookeeper.ConnectionLossException("Retry"))
c_d.addCallback(after_connected)
c_d.chainDeferred(d)
return d
# Dispatch from node watches on session expiration
def _watch_session_wrapper(self, watcher, event_type, conn_state, path):
"""Watch wrapper that diverts session events to a connection callback.
"""
if (event_type == zookeeper.SESSION_EVENT and
conn_state == zookeeper.EXPIRED_SESSION_STATE):
if self.unrecoverable:
log.debug("Watch got session expired event, reconnecting...")
d = self.cb_restablish_session()
d.addErrback(self._cb_restablish_errback)
return d
if event_type == zookeeper.SESSION_EVENT:
if self._session_event_callback:
self._session_event_callback(
self, ClientEvent(
event_type, conn_state, path, self.handle))
else:
return watcher(event_type, conn_state, path)
# Track all watches
def _wrap_watcher(self, watcher, watch_type, path):
if watcher is None:
return watcher
if not callable(watcher):
raise SyntaxError("invalid watcher")
# handle conn watcher, separately.
if watch_type is None and path is None:
return self._zk_thread_callback(
self._watch_session_wrapper, watcher)
return self._zk_thread_callback(
partial(
self._watch_session_wrapper,
self._watches.add(path, watch_type, watcher)))
# Track ephemerals
def _cb_created(self, d, data, acls, flags, result_code, path):
if self._check_result(result_code, d, path=path):
return
if (flags & zookeeper.EPHEMERAL) and not (flags & zookeeper.SEQUENCE):
self._ephemerals[path] = dict(
data=data, acls=acls, flags=flags)
d.callback(path)
def _cb_deleted(self, d, path, result_code):
if self._check_result(result_code, d, path=path):
return
self._ephemerals.pop(path, None)
d.callback(result_code)
def _cb_set_acl(self, d, path, acls, result_code):
if self._check_result(result_code, d, path=path):
return
if path in self._ephemerals:
self._ephemerals[path]['acls'] = acls
d.callback(result_code)
def _cb_set(self, d, path, data, result_code, node_stat):
if self._check_result(result_code, d, path=path):
return
if path in self._ephemerals:
self._ephemerals[path]['data'] = data
d.callback(node_stat)
class _ManagedClient(RetryClient):
def subscribe_new_session(self):
return self.client.subscribe_new_session()
def ManagedClient(servers=None, session_timeout=None, connect_timeout=10000):
client = SessionClient(servers, session_timeout, connect_timeout)
managed_client = _ManagedClient(client)
managed_client.set_retry_error_callback(client.cb_retry_expired)
return managed_client
|