root / daemons / ganeti-masterd @ c979d253
History | View | Annotate | Download (13.9 kB)
1 |
#!/usr/bin/python -u |
---|---|
2 |
# |
3 |
|
4 |
# Copyright (C) 2006, 2007 Google Inc. |
5 |
# |
6 |
# This program is free software; you can redistribute it and/or modify |
7 |
# it under the terms of the GNU General Public License as published by |
8 |
# the Free Software Foundation; either version 2 of the License, or |
9 |
# (at your option) any later version. |
10 |
# |
11 |
# This program is distributed in the hope that it will be useful, but |
12 |
# WITHOUT ANY WARRANTY; without even the implied warranty of |
13 |
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
14 |
# General Public License for more details. |
15 |
# |
16 |
# You should have received a copy of the GNU General Public License |
17 |
# along with this program; if not, write to the Free Software |
18 |
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA |
19 |
# 02110-1301, USA. |
20 |
|
21 |
|
22 |
"""Master daemon program. |
23 |
|
24 |
Some classes deviates from the standard style guide since the |
25 |
inheritance from parent classes requires it. |
26 |
|
27 |
""" |
28 |
|
29 |
|
30 |
import os |
31 |
import errno |
32 |
import sys |
33 |
import SocketServer |
34 |
import time |
35 |
import collections |
36 |
import Queue |
37 |
import random |
38 |
import signal |
39 |
import simplejson |
40 |
import logging |
41 |
|
42 |
from cStringIO import StringIO |
43 |
from optparse import OptionParser |
44 |
|
45 |
from ganeti import config |
46 |
from ganeti import constants |
47 |
from ganeti import mcpu |
48 |
from ganeti import opcodes |
49 |
from ganeti import jqueue |
50 |
from ganeti import locking |
51 |
from ganeti import luxi |
52 |
from ganeti import utils |
53 |
from ganeti import errors |
54 |
from ganeti import ssconf |
55 |
from ganeti import workerpool |
56 |
from ganeti import rpc |
57 |
from ganeti import bootstrap |
58 |
|
59 |
|
60 |
CLIENT_REQUEST_WORKERS = 16 |
61 |
|
62 |
EXIT_NOTMASTER = constants.EXIT_NOTMASTER |
63 |
EXIT_NODESETUP_ERROR = constants.EXIT_NODESETUP_ERROR |
64 |
|
65 |
|
66 |
class ClientRequestWorker(workerpool.BaseWorker): |
67 |
def RunTask(self, server, request, client_address): |
68 |
"""Process the request. |
69 |
|
70 |
This is copied from the code in ThreadingMixIn. |
71 |
|
72 |
""" |
73 |
try: |
74 |
server.finish_request(request, client_address) |
75 |
server.close_request(request) |
76 |
except: |
77 |
server.handle_error(request, client_address) |
78 |
server.close_request(request) |
79 |
|
80 |
|
81 |
class IOServer(SocketServer.UnixStreamServer): |
82 |
"""IO thread class. |
83 |
|
84 |
This class takes care of initializing the other threads, setting |
85 |
signal handlers (which are processed only in this thread), and doing |
86 |
cleanup at shutdown. |
87 |
|
88 |
""" |
89 |
def __init__(self, address, rqhandler): |
90 |
"""IOServer constructor |
91 |
|
92 |
@param address: the address to bind this IOServer to |
93 |
@param rqhandler: RequestHandler type object |
94 |
|
95 |
""" |
96 |
SocketServer.UnixStreamServer.__init__(self, address, rqhandler) |
97 |
|
98 |
# We'll only start threads once we've forked. |
99 |
self.context = None |
100 |
self.request_workers = None |
101 |
|
102 |
def setup_queue(self): |
103 |
self.context = GanetiContext() |
104 |
self.request_workers = workerpool.WorkerPool(CLIENT_REQUEST_WORKERS, |
105 |
ClientRequestWorker) |
106 |
|
107 |
def process_request(self, request, client_address): |
108 |
"""Add task to workerpool to process request. |
109 |
|
110 |
""" |
111 |
self.request_workers.AddTask(self, request, client_address) |
112 |
|
113 |
def serve_forever(self): |
114 |
"""Handle one request at a time until told to quit.""" |
115 |
sighandler = utils.SignalHandler([signal.SIGINT, signal.SIGTERM]) |
116 |
try: |
117 |
while not sighandler.called: |
118 |
self.handle_request() |
119 |
finally: |
120 |
sighandler.Reset() |
121 |
|
122 |
def server_cleanup(self): |
123 |
"""Cleanup the server. |
124 |
|
125 |
This involves shutting down the processor threads and the master |
126 |
socket. |
127 |
|
128 |
""" |
129 |
try: |
130 |
self.server_close() |
131 |
finally: |
132 |
if self.request_workers: |
133 |
self.request_workers.TerminateWorkers() |
134 |
if self.context: |
135 |
self.context.jobqueue.Shutdown() |
136 |
|
137 |
|
138 |
class ClientRqHandler(SocketServer.BaseRequestHandler): |
139 |
"""Client handler""" |
140 |
EOM = '\3' |
141 |
READ_SIZE = 4096 |
142 |
|
143 |
def setup(self): |
144 |
self._buffer = "" |
145 |
self._msgs = collections.deque() |
146 |
self._ops = ClientOps(self.server) |
147 |
|
148 |
def handle(self): |
149 |
while True: |
150 |
msg = self.read_message() |
151 |
if msg is None: |
152 |
logging.debug("client closed connection") |
153 |
break |
154 |
|
155 |
request = simplejson.loads(msg) |
156 |
logging.debug("request: %s", request) |
157 |
if not isinstance(request, dict): |
158 |
logging.error("wrong request received: %s", msg) |
159 |
break |
160 |
|
161 |
method = request.get(luxi.KEY_METHOD, None) |
162 |
args = request.get(luxi.KEY_ARGS, None) |
163 |
if method is None or args is None: |
164 |
logging.error("no method or args in request") |
165 |
break |
166 |
|
167 |
success = False |
168 |
try: |
169 |
result = self._ops.handle_request(method, args) |
170 |
success = True |
171 |
except errors.GenericError, err: |
172 |
success = False |
173 |
result = (err.__class__.__name__, err.args) |
174 |
except: |
175 |
logging.error("Unexpected exception", exc_info=True) |
176 |
err = sys.exc_info() |
177 |
result = "Caught exception: %s" % str(err[1]) |
178 |
|
179 |
response = { |
180 |
luxi.KEY_SUCCESS: success, |
181 |
luxi.KEY_RESULT: result, |
182 |
} |
183 |
logging.debug("response: %s", response) |
184 |
self.send_message(simplejson.dumps(response)) |
185 |
|
186 |
def read_message(self): |
187 |
while not self._msgs: |
188 |
data = self.request.recv(self.READ_SIZE) |
189 |
if not data: |
190 |
return None |
191 |
new_msgs = (self._buffer + data).split(self.EOM) |
192 |
self._buffer = new_msgs.pop() |
193 |
self._msgs.extend(new_msgs) |
194 |
return self._msgs.popleft() |
195 |
|
196 |
def send_message(self, msg): |
197 |
#print "sending", msg |
198 |
self.request.sendall(msg + self.EOM) |
199 |
|
200 |
|
201 |
class ClientOps: |
202 |
"""Class holding high-level client operations.""" |
203 |
def __init__(self, server): |
204 |
self.server = server |
205 |
|
206 |
def handle_request(self, method, args): |
207 |
queue = self.server.context.jobqueue |
208 |
|
209 |
# TODO: Parameter validation |
210 |
|
211 |
if method == luxi.REQ_SUBMIT_JOB: |
212 |
ops = [opcodes.OpCode.LoadOpCode(state) for state in args] |
213 |
return queue.SubmitJob(ops) |
214 |
|
215 |
elif method == luxi.REQ_CANCEL_JOB: |
216 |
job_id = args |
217 |
return queue.CancelJob(job_id) |
218 |
|
219 |
elif method == luxi.REQ_ARCHIVE_JOB: |
220 |
job_id = args |
221 |
return queue.ArchiveJob(job_id) |
222 |
|
223 |
elif method == luxi.REQ_AUTOARCHIVE_JOBS: |
224 |
(age, timeout) = args |
225 |
return queue.AutoArchiveJobs(age, timeout) |
226 |
|
227 |
elif method == luxi.REQ_WAIT_FOR_JOB_CHANGE: |
228 |
(job_id, fields, prev_job_info, prev_log_serial, timeout) = args |
229 |
return queue.WaitForJobChanges(job_id, fields, prev_job_info, |
230 |
prev_log_serial, timeout) |
231 |
|
232 |
elif method == luxi.REQ_QUERY_JOBS: |
233 |
(job_ids, fields) = args |
234 |
return queue.QueryJobs(job_ids, fields) |
235 |
|
236 |
elif method == luxi.REQ_QUERY_INSTANCES: |
237 |
(names, fields) = args |
238 |
op = opcodes.OpQueryInstances(names=names, output_fields=fields) |
239 |
return self._Query(op) |
240 |
|
241 |
elif method == luxi.REQ_QUERY_NODES: |
242 |
(names, fields) = args |
243 |
op = opcodes.OpQueryNodes(names=names, output_fields=fields) |
244 |
return self._Query(op) |
245 |
|
246 |
elif method == luxi.REQ_QUERY_EXPORTS: |
247 |
nodes = args |
248 |
op = opcodes.OpQueryExports(nodes=nodes) |
249 |
return self._Query(op) |
250 |
|
251 |
elif method == luxi.REQ_QUERY_CONFIG_VALUES: |
252 |
fields = args |
253 |
op = opcodes.OpQueryConfigValues(output_fields=fields) |
254 |
return self._Query(op) |
255 |
|
256 |
elif method == luxi.REQ_QUEUE_SET_DRAIN_FLAG: |
257 |
drain_flag = args |
258 |
return queue.SetDrainFlag(drain_flag) |
259 |
|
260 |
else: |
261 |
raise ValueError("Invalid operation") |
262 |
|
263 |
def _DummyLog(self, *args): |
264 |
pass |
265 |
|
266 |
def _Query(self, op): |
267 |
"""Runs the specified opcode and returns the result. |
268 |
|
269 |
""" |
270 |
proc = mcpu.Processor(self.server.context) |
271 |
# TODO: Where should log messages go? |
272 |
return proc.ExecOpCode(op, self._DummyLog, None) |
273 |
|
274 |
|
275 |
class GanetiContext(object): |
276 |
"""Context common to all ganeti threads. |
277 |
|
278 |
This class creates and holds common objects shared by all threads. |
279 |
|
280 |
""" |
281 |
_instance = None |
282 |
|
283 |
def __init__(self): |
284 |
"""Constructs a new GanetiContext object. |
285 |
|
286 |
There should be only a GanetiContext object at any time, so this |
287 |
function raises an error if this is not the case. |
288 |
|
289 |
""" |
290 |
assert self.__class__._instance is None, "double GanetiContext instance" |
291 |
|
292 |
# Create global configuration object |
293 |
self.cfg = config.ConfigWriter() |
294 |
|
295 |
# Locking manager |
296 |
self.glm = locking.GanetiLockManager( |
297 |
self.cfg.GetNodeList(), |
298 |
self.cfg.GetInstanceList()) |
299 |
|
300 |
# Job queue |
301 |
self.jobqueue = jqueue.JobQueue(self) |
302 |
|
303 |
# setting this also locks the class against attribute modifications |
304 |
self.__class__._instance = self |
305 |
|
306 |
def __setattr__(self, name, value): |
307 |
"""Setting GanetiContext attributes is forbidden after initialization. |
308 |
|
309 |
""" |
310 |
assert self.__class__._instance is None, "Attempt to modify Ganeti Context" |
311 |
object.__setattr__(self, name, value) |
312 |
|
313 |
def AddNode(self, node): |
314 |
"""Adds a node to the configuration and lock manager. |
315 |
|
316 |
""" |
317 |
# Add it to the configuration |
318 |
self.cfg.AddNode(node) |
319 |
|
320 |
# If preseeding fails it'll not be added |
321 |
self.jobqueue.AddNode(node) |
322 |
|
323 |
# Add the new node to the Ganeti Lock Manager |
324 |
self.glm.add(locking.LEVEL_NODE, node.name) |
325 |
|
326 |
def ReaddNode(self, node): |
327 |
"""Updates a node that's already in the configuration |
328 |
|
329 |
""" |
330 |
# Synchronize the queue again |
331 |
self.jobqueue.AddNode(node) |
332 |
|
333 |
def RemoveNode(self, name): |
334 |
"""Removes a node from the configuration and lock manager. |
335 |
|
336 |
""" |
337 |
# Remove node from configuration |
338 |
self.cfg.RemoveNode(name) |
339 |
|
340 |
# Notify job queue |
341 |
self.jobqueue.RemoveNode(name) |
342 |
|
343 |
# Remove the node from the Ganeti Lock Manager |
344 |
self.glm.remove(locking.LEVEL_NODE, name) |
345 |
|
346 |
|
347 |
def ParseOptions(): |
348 |
"""Parse the command line options. |
349 |
|
350 |
@return: (options, args) as from OptionParser.parse_args() |
351 |
|
352 |
""" |
353 |
parser = OptionParser(description="Ganeti master daemon", |
354 |
usage="%prog [-f] [-d]", |
355 |
version="%%prog (ganeti) %s" % |
356 |
constants.RELEASE_VERSION) |
357 |
|
358 |
parser.add_option("-f", "--foreground", dest="fork", |
359 |
help="Don't detach from the current terminal", |
360 |
default=True, action="store_false") |
361 |
parser.add_option("-d", "--debug", dest="debug", |
362 |
help="Enable some debug messages", |
363 |
default=False, action="store_true") |
364 |
options, args = parser.parse_args() |
365 |
return options, args |
366 |
|
367 |
|
368 |
def CheckAgreement(): |
369 |
"""Check the agreement on who is the master. |
370 |
|
371 |
The function uses a very simple algorithm: we must get more positive |
372 |
than negative answers. Since in most of the cases we are the master, |
373 |
we'll use our own config file for getting the node list. In the |
374 |
future we could collect the current node list from our (possibly |
375 |
obsolete) known nodes. |
376 |
|
377 |
In order to account for cold-start of all nodes, we retry for up to |
378 |
a minute until we get a real answer as the top-voted one. If the |
379 |
nodes are more out-of-sync, for now manual startup of the master |
380 |
should be attempted. |
381 |
|
382 |
Note that for a even number of nodes cluster, we need at least half |
383 |
of the nodes (beside ourselves) to vote for us. This creates a |
384 |
problem on two-node clusters, since in this case we require the |
385 |
other node to be up too to confirm our status. |
386 |
|
387 |
""" |
388 |
myself = utils.HostInfo().name |
389 |
#temp instantiation of a config writer, used only to get the node list |
390 |
cfg = config.ConfigWriter() |
391 |
node_list = cfg.GetNodeList() |
392 |
del cfg |
393 |
retries = 6 |
394 |
while retries > 0: |
395 |
votes = bootstrap.GatherMasterVotes(node_list) |
396 |
if not votes: |
397 |
# empty node list, this is a one node cluster |
398 |
return True |
399 |
if votes[0][0] is None: |
400 |
retries -= 1 |
401 |
time.sleep(10) |
402 |
continue |
403 |
break |
404 |
if retries == 0: |
405 |
logging.critical("Cluster inconsistent, most of the nodes didn't answer" |
406 |
" after multiple retries. Aborting startup") |
407 |
return False |
408 |
# here a real node is at the top of the list |
409 |
all_votes = sum(item[1] for item in votes) |
410 |
top_node, top_votes = votes[0] |
411 |
result = False |
412 |
if top_node != myself: |
413 |
logging.critical("It seems we are not the master (top-voted node" |
414 |
" is %s with %d out of %d votes)", top_node, top_votes, |
415 |
all_votes) |
416 |
elif top_votes < all_votes - top_votes: |
417 |
logging.critical("It seems we are not the master (%d votes for," |
418 |
" %d votes against)", top_votes, all_votes - top_votes) |
419 |
else: |
420 |
result = True |
421 |
|
422 |
return result |
423 |
|
424 |
|
425 |
def main(): |
426 |
"""Main function""" |
427 |
|
428 |
options, args = ParseOptions() |
429 |
utils.debug = options.debug |
430 |
utils.no_fork = True |
431 |
|
432 |
if options.fork: |
433 |
utils.CloseFDs() |
434 |
|
435 |
rpc.Init() |
436 |
try: |
437 |
ssconf.CheckMaster(options.debug) |
438 |
|
439 |
# we believe we are the master, let's ask the other nodes... |
440 |
if not CheckAgreement(): |
441 |
return |
442 |
|
443 |
dirs = [(constants.RUN_GANETI_DIR, constants.RUN_DIRS_MODE), |
444 |
(constants.SOCKET_DIR, constants.SOCKET_DIR_MODE), |
445 |
] |
446 |
for dir_name, mode in dirs: |
447 |
try: |
448 |
os.mkdir(dir_name, mode) |
449 |
except EnvironmentError, err: |
450 |
if err.errno != errno.EEXIST: |
451 |
raise errors.GenericError("Cannot create needed directory" |
452 |
" '%s': %s" % (constants.SOCKET_DIR, err)) |
453 |
if not os.path.isdir(dir_name): |
454 |
raise errors.GenericError("%s is not a directory" % dir_name) |
455 |
|
456 |
# This is safe to do as the pid file guarantees against |
457 |
# concurrent execution. |
458 |
utils.RemoveFile(constants.MASTER_SOCKET) |
459 |
|
460 |
master = IOServer(constants.MASTER_SOCKET, ClientRqHandler) |
461 |
finally: |
462 |
rpc.Shutdown() |
463 |
|
464 |
# become a daemon |
465 |
if options.fork: |
466 |
utils.Daemonize(logfile=constants.LOG_MASTERDAEMON) |
467 |
|
468 |
utils.WritePidFile(constants.MASTERD_PID) |
469 |
try: |
470 |
utils.SetupLogging(constants.LOG_MASTERDAEMON, debug=options.debug, |
471 |
stderr_logging=not options.fork, multithreaded=True) |
472 |
|
473 |
logging.info("Ganeti master daemon startup") |
474 |
|
475 |
rpc.Init() |
476 |
try: |
477 |
# activate ip |
478 |
master_node = ssconf.SimpleConfigReader().GetMasterNode() |
479 |
if not rpc.RpcRunner.call_node_start_master(master_node, False): |
480 |
logging.error("Can't activate master IP address") |
481 |
|
482 |
master.setup_queue() |
483 |
try: |
484 |
master.serve_forever() |
485 |
finally: |
486 |
master.server_cleanup() |
487 |
finally: |
488 |
rpc.Shutdown() |
489 |
finally: |
490 |
utils.RemovePidFile(constants.MASTERD_PID) |
491 |
utils.RemoveFile(constants.MASTER_SOCKET) |
492 |
|
493 |
|
494 |
if __name__ == "__main__": |
495 |
main() |