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
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
|
#!/usr/bin/env python
# -*- coding: utf-8 -*-
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements. See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership. The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License. You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import with_statement
import logging
import os
import random
import shutil
import subprocess
import sys
import tempfile
import time
import urllib2
from optparse import OptionParser
from sys import stderr
import boto
from boto.ec2.blockdevicemapping import BlockDeviceMapping, EBSBlockDeviceType
from boto import ec2
# A URL prefix from which to fetch AMI information
AMI_PREFIX = "https://raw.github.com/pwendell/spark-ec2/ec2-updates/ami-list"
# Configure and parse our command-line arguments
def parse_args():
parser = OptionParser(usage="spark-ec2 [options] <action> <cluster_name>"
+ "\n\n<action> can be: launch, destroy, login, stop, start, get-master",
add_help_option=False)
parser.add_option("-h", "--help", action="help",
help="Show this help message and exit")
parser.add_option("-s", "--slaves", type="int", default=1,
help="Number of slaves to launch (default: 1)")
parser.add_option("-w", "--wait", type="int", default=120,
help="Seconds to wait for nodes to start (default: 120)")
parser.add_option("-k", "--key-pair",
help="Key pair to use on instances")
parser.add_option("-i", "--identity-file",
help="SSH private key file to use for logging into instances")
parser.add_option("-t", "--instance-type", default="m1.large",
help="Type of instance to launch (default: m1.large). " +
"WARNING: must be 64-bit; small instances won't work")
parser.add_option("-m", "--master-instance-type", default="",
help="Master instance type (leave empty for same as instance-type)")
parser.add_option("-r", "--region", default="us-east-1",
help="EC2 region zone to launch instances in")
parser.add_option("-z", "--zone", default="",
help="Availability zone to launch instances in, or 'all' to spread " +
"slaves across multiple (an additional $0.01/Gb for bandwidth" +
"between zones applies)")
parser.add_option("-a", "--ami", default="v0.7.0",
help="Amazon Machine Image ID to use, or 'vX.Y.Z' to use version " +
"X.Y.Z of Spark (default: v0.7.0)")
parser.add_option("-D", metavar="[ADDRESS:]PORT", dest="proxy_port",
help="Use SSH dynamic port forwarding to create a SOCKS proxy at " +
"the given local address (for use with login)")
parser.add_option("--resume", action="store_true", default=False,
help="Resume installation on a previously launched cluster " +
"(for debugging)")
parser.add_option("--ebs-vol-size", metavar="SIZE", type="int", default=0,
help="Attach a new EBS volume of size SIZE (in GB) to each node as " +
"/vol. The volumes will be deleted when the instances terminate. " +
"Only possible on EBS-backed AMIs.")
parser.add_option("--swap", metavar="SWAP", type="int", default=1024,
help="Swap space to set up per node, in MB (default: 1024)")
parser.add_option("--spot-price", metavar="PRICE", type="float",
help="If specified, launch slaves as spot instances with the given " +
"maximum price (in dollars)")
parser.add_option("--cluster-type", type="choice", metavar="TYPE",
choices=["mesos", "standalone"], default="standalone",
help="'mesos' for a Mesos cluster, 'standalone' for a standalone " +
"Spark cluster (default: standalone)")
parser.add_option("--ganglia", action="store_true", default=True,
help="Setup Ganglia monitoring on cluster (default: on). NOTE: " +
"the Ganglia page will be publicly accessible")
parser.add_option("--no-ganglia", action="store_false", dest="ganglia",
help="Disable Ganglia monitoring for the cluster")
parser.add_option("--old-scripts", action="store_true", default=False,
help="Use old mesos-ec2 scripts, for Spark <= 0.6 AMIs")
parser.add_option("-u", "--user", default="root",
help="The SSH user you want to connect as (default: root)")
parser.add_option("--delete-groups", action="store_true", default=False,
help="When destroying a cluster, delete the security groups that were created")
(opts, args) = parser.parse_args()
if len(args) != 2:
parser.print_help()
sys.exit(1)
(action, cluster_name) = args
if opts.identity_file == None and action in ['launch', 'login']:
print >> stderr, ("ERROR: The -i or --identity-file argument is " +
"required for " + action)
sys.exit(1)
if opts.cluster_type not in ["mesos", "standalone"] and action == "launch":
print >> stderr, ("ERROR: Invalid cluster type: " + opts.cluster_type)
sys.exit(1)
# Boto config check
# http://boto.cloudhackers.com/en/latest/boto_config_tut.html
home_dir = os.getenv('HOME')
if home_dir == None or not os.path.isfile(home_dir + '/.boto'):
if not os.path.isfile('/etc/boto.cfg'):
if os.getenv('AWS_ACCESS_KEY_ID') == None:
print >> stderr, ("ERROR: The environment variable AWS_ACCESS_KEY_ID " +
"must be set")
sys.exit(1)
if os.getenv('AWS_SECRET_ACCESS_KEY') == None:
print >> stderr, ("ERROR: The environment variable AWS_SECRET_ACCESS_KEY " +
"must be set")
sys.exit(1)
return (opts, action, cluster_name)
# Get the EC2 security group of the given name, creating it if it doesn't exist
def get_or_make_group(conn, name):
groups = conn.get_all_security_groups()
group = [g for g in groups if g.name == name]
if len(group) > 0:
return group[0]
else:
print "Creating security group " + name
return conn.create_security_group(name, "Spark EC2 group")
# Wait for a set of launched instances to exit the "pending" state
# (i.e. either to start running or to fail and be terminated)
def wait_for_instances(conn, instances):
while True:
for i in instances:
i.update()
if len([i for i in instances if i.state == 'pending']) > 0:
time.sleep(5)
else:
return
# Check whether a given EC2 instance object is in a state we consider active,
# i.e. not terminating or terminated. We count both stopping and stopped as
# active since we can restart stopped clusters.
def is_active(instance):
return (instance.state in ['pending', 'running', 'stopping', 'stopped'])
# Attempt to resolve an appropriate AMI given the architecture and
# region of the request.
def get_spark_ami(opts):
version = opts.ami
instance_types = {
"m1.small": "pvm",
"m1.medium": "pvm",
"m1.large": "pvm",
"m1.xlarge": "pvm",
"t1.micro": "pvm",
"c1.medium": "pvm",
"c1.xlarge": "pvm",
"m2.xlarge": "pvm",
"m2.2xlarge": "pvm",
"m2.4xlarge": "pvm",
"cc1.4xlarge": "hvm",
"cc2.8xlarge": "hvm",
"cg1.4xlarge": "hvm",
"hs1.8xlarge": "hvm",
"hi1.4xlarge": "hvm",
"m3.xlarge": "hvm",
"m3.2xlarge": "hvm",
"cr1.8xlarge": "hvm"
}
if opts.instance_type in instance_types:
instance_type = instance_types[opts.instance_type]
else:
instance_type = "pvm"
print >> stderr,\
"Don't recognize %s, assuming type is pvm" % opts.instance_type
if version != "v0.7.0":
print >> stderr, \
"Don't know how to resolve AMI for version: %s" % version
# TODO(pwendell) Once we have multiple Spark AMI versions, we should let
# people give a version flag here in place of just saying 'latest'.
version = version[1:]
ami_path = "%s/%s/%s/%s" % (AMI_PREFIX, version, opts.region, instance_type)
try:
ami = urllib2.urlopen(ami_path).read().strip()
print "Spark AMI: " + ami
except:
print >> stderr, "Could not resolve AMI at: " + ami_path
sys.exit(1)
return ami
# Launch a cluster of the given name, by setting up its security groups,
# and then starting new instances in them.
# Returns a tuple of EC2 reservation objects for the master, slave
# and zookeeper instances (in that order).
# Fails if there already instances running in the cluster's groups.
def launch_cluster(conn, opts, cluster_name):
print "Setting up security groups..."
master_group = get_or_make_group(conn, cluster_name + "-master")
slave_group = get_or_make_group(conn, cluster_name + "-slaves")
zoo_group = get_or_make_group(conn, cluster_name + "-zoo")
if master_group.rules == []: # Group was just now created
master_group.authorize(src_group=master_group)
master_group.authorize(src_group=slave_group)
master_group.authorize(src_group=zoo_group)
master_group.authorize('tcp', 22, 22, '0.0.0.0/0')
master_group.authorize('tcp', 8080, 8081, '0.0.0.0/0')
master_group.authorize('tcp', 50030, 50030, '0.0.0.0/0')
master_group.authorize('tcp', 50070, 50070, '0.0.0.0/0')
master_group.authorize('tcp', 60070, 60070, '0.0.0.0/0')
if opts.cluster_type == "mesos":
master_group.authorize('tcp', 38090, 38090, '0.0.0.0/0')
if opts.ganglia:
master_group.authorize('tcp', 5080, 5080, '0.0.0.0/0')
if slave_group.rules == []: # Group was just now created
slave_group.authorize(src_group=master_group)
slave_group.authorize(src_group=slave_group)
slave_group.authorize(src_group=zoo_group)
slave_group.authorize('tcp', 22, 22, '0.0.0.0/0')
slave_group.authorize('tcp', 8080, 8081, '0.0.0.0/0')
slave_group.authorize('tcp', 50060, 50060, '0.0.0.0/0')
slave_group.authorize('tcp', 50075, 50075, '0.0.0.0/0')
slave_group.authorize('tcp', 60060, 60060, '0.0.0.0/0')
slave_group.authorize('tcp', 60075, 60075, '0.0.0.0/0')
if zoo_group.rules == []: # Group was just now created
zoo_group.authorize(src_group=master_group)
zoo_group.authorize(src_group=slave_group)
zoo_group.authorize(src_group=zoo_group)
zoo_group.authorize('tcp', 22, 22, '0.0.0.0/0')
zoo_group.authorize('tcp', 2181, 2181, '0.0.0.0/0')
zoo_group.authorize('tcp', 2888, 2888, '0.0.0.0/0')
zoo_group.authorize('tcp', 3888, 3888, '0.0.0.0/0')
# Check if instances are already running in our groups
active_nodes = get_existing_cluster(conn, opts, cluster_name,
die_on_error=False)
if any(active_nodes):
print >> stderr, ("ERROR: There are already instances running in " +
"group %s, %s or %s" % (master_group.name, slave_group.name, zoo_group.name))
sys.exit(1)
# Figure out Spark AMI
if opts.ami[0] == "v":
opts.ami = get_spark_ami(opts)
print "Launching instances..."
try:
image = conn.get_all_images(image_ids=[opts.ami])[0]
except:
print >> stderr, "Could not find AMI " + opts.ami
sys.exit(1)
# Create block device mapping so that we can add an EBS volume if asked to
block_map = BlockDeviceMapping()
if opts.ebs_vol_size > 0:
device = EBSBlockDeviceType()
device.size = opts.ebs_vol_size
device.delete_on_termination = True
block_map["/dev/sdv"] = device
# Launch slaves
if opts.spot_price != None:
# Launch spot instances with the requested price
print ("Requesting %d slaves as spot instances with price $%.3f" %
(opts.slaves, opts.spot_price))
zones = get_zones(conn, opts)
num_zones = len(zones)
i = 0
my_req_ids = []
for zone in zones:
num_slaves_this_zone = get_partition(opts.slaves, num_zones, i)
slave_reqs = conn.request_spot_instances(
price = opts.spot_price,
image_id = opts.ami,
launch_group = "launch-group-%s" % cluster_name,
placement = zone,
count = num_slaves_this_zone,
key_name = opts.key_pair,
security_groups = [slave_group],
instance_type = opts.instance_type,
block_device_map = block_map)
my_req_ids += [req.id for req in slave_reqs]
i += 1
print "Waiting for spot instances to be granted..."
try:
while True:
time.sleep(10)
reqs = conn.get_all_spot_instance_requests()
id_to_req = {}
for r in reqs:
id_to_req[r.id] = r
active_instance_ids = []
for i in my_req_ids:
if i in id_to_req and id_to_req[i].state == "active":
active_instance_ids.append(id_to_req[i].instance_id)
if len(active_instance_ids) == opts.slaves:
print "All %d slaves granted" % opts.slaves
reservations = conn.get_all_instances(active_instance_ids)
slave_nodes = []
for r in reservations:
slave_nodes += r.instances
break
else:
print "%d of %d slaves granted, waiting longer" % (
len(active_instance_ids), opts.slaves)
except:
print "Canceling spot instance requests"
conn.cancel_spot_instance_requests(my_req_ids)
# Log a warning if any of these requests actually launched instances:
(master_nodes, slave_nodes, zoo_nodes) = get_existing_cluster(
conn, opts, cluster_name, die_on_error=False)
running = len(master_nodes) + len(slave_nodes) + len(zoo_nodes)
if running:
print >> stderr, ("WARNING: %d instances are still running" % running)
sys.exit(0)
else:
# Launch non-spot instances
zones = get_zones(conn, opts)
num_zones = len(zones)
i = 0
slave_nodes = []
for zone in zones:
num_slaves_this_zone = get_partition(opts.slaves, num_zones, i)
if num_slaves_this_zone > 0:
slave_res = image.run(key_name = opts.key_pair,
security_groups = [slave_group],
instance_type = opts.instance_type,
placement = zone,
min_count = num_slaves_this_zone,
max_count = num_slaves_this_zone,
block_device_map = block_map)
slave_nodes += slave_res.instances
print "Launched %d slaves in %s, regid = %s" % (num_slaves_this_zone,
zone, slave_res.id)
i += 1
# Launch masters
master_type = opts.master_instance_type
if master_type == "":
master_type = opts.instance_type
if opts.zone == 'all':
opts.zone = random.choice(conn.get_all_zones()).name
master_res = image.run(key_name = opts.key_pair,
security_groups = [master_group],
instance_type = master_type,
placement = opts.zone,
min_count = 1,
max_count = 1,
block_device_map = block_map)
master_nodes = master_res.instances
print "Launched master in %s, regid = %s" % (zone, master_res.id)
zoo_nodes = []
# Return all the instances
return (master_nodes, slave_nodes, zoo_nodes)
# Get the EC2 instances in an existing cluster if available.
# Returns a tuple of lists of EC2 instance objects for the masters,
# slaves and zookeeper nodes (in that order).
def get_existing_cluster(conn, opts, cluster_name, die_on_error=True):
print "Searching for existing cluster " + cluster_name + "..."
reservations = conn.get_all_instances()
master_nodes = []
slave_nodes = []
zoo_nodes = []
for res in reservations:
active = [i for i in res.instances if is_active(i)]
if len(active) > 0:
group_names = [g.name for g in res.groups]
if group_names == [cluster_name + "-master"]:
master_nodes += res.instances
elif group_names == [cluster_name + "-slaves"]:
slave_nodes += res.instances
elif group_names == [cluster_name + "-zoo"]:
zoo_nodes += res.instances
if any((master_nodes, slave_nodes, zoo_nodes)):
print ("Found %d master(s), %d slaves, %d ZooKeeper nodes" %
(len(master_nodes), len(slave_nodes), len(zoo_nodes)))
if (master_nodes != [] and slave_nodes != []) or not die_on_error:
return (master_nodes, slave_nodes, zoo_nodes)
else:
if master_nodes == [] and slave_nodes != []:
print "ERROR: Could not find master in group " + cluster_name + "-master"
elif master_nodes != [] and slave_nodes == []:
print "ERROR: Could not find slaves in group " + cluster_name + "-slaves"
else:
print "ERROR: Could not find any existing cluster"
sys.exit(1)
# Deploy configuration files and run setup scripts on a newly launched
# or started EC2 cluster.
def setup_cluster(conn, master_nodes, slave_nodes, zoo_nodes, opts, deploy_ssh_key):
master = master_nodes[0].public_dns_name
if deploy_ssh_key:
print "Copying SSH key %s to master..." % opts.identity_file
ssh(master, opts, 'mkdir -p ~/.ssh')
scp(master, opts, opts.identity_file, '~/.ssh/id_rsa')
ssh(master, opts, 'chmod 600 ~/.ssh/id_rsa')
if opts.cluster_type == "mesos":
modules = ['ephemeral-hdfs', 'persistent-hdfs', 'mapreduce', 'mesos']
elif opts.cluster_type == "standalone":
modules = ['ephemeral-hdfs', 'persistent-hdfs', 'mapreduce',
'spark-standalone']
if opts.ganglia:
modules.append('ganglia')
if not opts.old_scripts:
# NOTE: We should clone the repository before running deploy_files to
# prevent ec2-variables.sh from being overwritten
ssh(master, opts, "rm -rf spark-ec2 && git clone https://github.com/pwendell/spark-ec2.git -b ec2-updates")
print "Deploying files to master..."
deploy_files(conn, "deploy.generic", opts, master_nodes, slave_nodes,
zoo_nodes, modules)
print "Running setup on master..."
if opts.old_scripts:
if opts.cluster_type == "mesos":
setup_mesos_cluster(master, opts)
elif opts.cluster_type == "standalone":
setup_standalone_cluster(master, slave_nodes, opts)
else:
setup_spark_cluster(master, opts)
print "Done!"
def setup_mesos_cluster(master, opts):
ssh(master, opts, "chmod u+x mesos-ec2/setup")
ssh(master, opts, "mesos-ec2/setup %s %s %s %s" %
("generic", "none", "master", opts.swap))
def setup_standalone_cluster(master, slave_nodes, opts):
slave_ips = '\n'.join([i.public_dns_name for i in slave_nodes])
ssh(master, opts, "echo \"%s\" > spark/conf/slaves" % (slave_ips))
ssh(master, opts, "/root/spark/bin/start-all.sh")
def setup_spark_cluster(master, opts):
ssh(master, opts, "chmod u+x spark-ec2/setup.sh")
ssh(master, opts, "spark-ec2/setup.sh")
if opts.cluster_type == "mesos":
print "Mesos cluster started at http://%s:8080" % master
elif opts.cluster_type == "standalone":
print "Spark standalone cluster started at http://%s:8080" % master
if opts.ganglia:
print "Ganglia started at http://%s:5080/ganglia" % master
# Wait for a whole cluster (masters, slaves and ZooKeeper) to start up
def wait_for_cluster(conn, wait_secs, master_nodes, slave_nodes, zoo_nodes):
print "Waiting for instances to start up..."
time.sleep(5)
wait_for_instances(conn, master_nodes)
wait_for_instances(conn, slave_nodes)
if zoo_nodes != []:
wait_for_instances(conn, zoo_nodes)
print "Waiting %d more seconds..." % wait_secs
time.sleep(wait_secs)
# Get number of local disks available for a given EC2 instance type.
def get_num_disks(instance_type):
# From http://docs.amazonwebservices.com/AWSEC2/latest/UserGuide/index.html?InstanceStorage.html
disks_by_instance = {
"m1.small": 1,
"m1.medium": 1,
"m1.large": 2,
"m1.xlarge": 4,
"t1.micro": 1,
"c1.medium": 1,
"c1.xlarge": 4,
"m2.xlarge": 1,
"m2.2xlarge": 1,
"m2.4xlarge": 2,
"cc1.4xlarge": 2,
"cc2.8xlarge": 4,
"cg1.4xlarge": 2,
"hs1.8xlarge": 24,
"cr1.8xlarge": 2,
"hi1.4xlarge": 2,
"m3.xlarge": 0,
"m3.2xlarge": 0
}
if instance_type in disks_by_instance:
return disks_by_instance[instance_type]
else:
print >> stderr, ("WARNING: Don't know number of disks on instance type %s; assuming 1"
% instance_type)
return 1
# Deploy the configuration file templates in a given local directory to
# a cluster, filling in any template parameters with information about the
# cluster (e.g. lists of masters and slaves). Files are only deployed to
# the first master instance in the cluster, and we expect the setup
# script to be run on that instance to copy them to other nodes.
def deploy_files(conn, root_dir, opts, master_nodes, slave_nodes, zoo_nodes,
modules):
active_master = master_nodes[0].public_dns_name
num_disks = get_num_disks(opts.instance_type)
hdfs_data_dirs = "/mnt/ephemeral-hdfs/data"
mapred_local_dirs = "/mnt/hadoop/mrlocal"
spark_local_dirs = "/mnt/spark"
if num_disks > 1:
for i in range(2, num_disks + 1):
hdfs_data_dirs += ",/mnt%d/ephemeral-hdfs/data" % i
mapred_local_dirs += ",/mnt%d/hadoop/mrlocal" % i
spark_local_dirs += ",/mnt%d/spark" % i
if zoo_nodes != []:
zoo_list = '\n'.join([i.public_dns_name for i in zoo_nodes])
cluster_url = "zoo://" + ",".join(
["%s:2181/mesos" % i.public_dns_name for i in zoo_nodes])
elif opts.cluster_type == "mesos":
zoo_list = "NONE"
cluster_url = "%s:5050" % active_master
elif opts.cluster_type == "standalone":
zoo_list = "NONE"
cluster_url = "%s:7077" % active_master
template_vars = {
"master_list": '\n'.join([i.public_dns_name for i in master_nodes]),
"active_master": active_master,
"slave_list": '\n'.join([i.public_dns_name for i in slave_nodes]),
"zoo_list": zoo_list,
"cluster_url": cluster_url,
"hdfs_data_dirs": hdfs_data_dirs,
"mapred_local_dirs": mapred_local_dirs,
"spark_local_dirs": spark_local_dirs,
"swap": str(opts.swap),
"modules": '\n'.join(modules)
}
# Create a temp directory in which we will place all the files to be
# deployed after we substitue template parameters in them
tmp_dir = tempfile.mkdtemp()
for path, dirs, files in os.walk(root_dir):
if path.find(".svn") == -1:
dest_dir = os.path.join('/', path[len(root_dir):])
local_dir = tmp_dir + dest_dir
if not os.path.exists(local_dir):
os.makedirs(local_dir)
for filename in files:
if filename[0] not in '#.~' and filename[-1] != '~':
dest_file = os.path.join(dest_dir, filename)
local_file = tmp_dir + dest_file
with open(os.path.join(path, filename)) as src:
with open(local_file, "w") as dest:
text = src.read()
for key in template_vars:
text = text.replace("{{" + key + "}}", template_vars[key])
dest.write(text)
dest.close()
# rsync the whole directory over to the master machine
command = (("rsync -rv -e 'ssh -o StrictHostKeyChecking=no -i %s' " +
"'%s/' '%s@%s:/'") % (opts.identity_file, tmp_dir, opts.user, active_master))
subprocess.check_call(command, shell=True)
# Remove the temp directory we created above
shutil.rmtree(tmp_dir)
# Copy a file to a given host through scp, throwing an exception if scp fails
def scp(host, opts, local_file, dest_file):
subprocess.check_call(
"scp -q -o StrictHostKeyChecking=no -i %s '%s' '%s@%s:%s'" %
(opts.identity_file, local_file, opts.user, host, dest_file), shell=True)
# Run a command on a host through ssh, retrying up to two times
# and then throwing an exception if ssh continues to fail.
def ssh(host, opts, command):
tries = 0
while True:
try:
return subprocess.check_call(
"ssh -t -o StrictHostKeyChecking=no -i %s %s@%s '%s'" %
(opts.identity_file, opts.user, host, command), shell=True)
except subprocess.CalledProcessError as e:
if (tries > 2):
raise e
print "Error connecting to host {0}, sleeping 30".format(e)
time.sleep(30)
tries = tries + 1
# Gets a list of zones to launch instances in
def get_zones(conn, opts):
if opts.zone == 'all':
zones = [z.name for z in conn.get_all_zones()]
else:
zones = [opts.zone]
return zones
# Gets the number of items in a partition
def get_partition(total, num_partitions, current_partitions):
num_slaves_this_zone = total / num_partitions
if (total % num_partitions) - current_partitions > 0:
num_slaves_this_zone += 1
return num_slaves_this_zone
def main():
(opts, action, cluster_name) = parse_args()
try:
conn = ec2.connect_to_region(opts.region)
except Exception as e:
print >> stderr, (e)
sys.exit(1)
# Select an AZ at random if it was not specified.
if opts.zone == "":
opts.zone = random.choice(conn.get_all_zones()).name
if action == "launch":
if opts.resume:
(master_nodes, slave_nodes, zoo_nodes) = get_existing_cluster(
conn, opts, cluster_name)
else:
(master_nodes, slave_nodes, zoo_nodes) = launch_cluster(
conn, opts, cluster_name)
wait_for_cluster(conn, opts.wait, master_nodes, slave_nodes, zoo_nodes)
setup_cluster(conn, master_nodes, slave_nodes, zoo_nodes, opts, True)
elif action == "destroy":
response = raw_input("Are you sure you want to destroy the cluster " +
cluster_name + "?\nALL DATA ON ALL NODES WILL BE LOST!!\n" +
"Destroy cluster " + cluster_name + " (y/N): ")
if response == "y":
(master_nodes, slave_nodes, zoo_nodes) = get_existing_cluster(
conn, opts, cluster_name, die_on_error=False)
print "Terminating master..."
for inst in master_nodes:
inst.terminate()
print "Terminating slaves..."
for inst in slave_nodes:
inst.terminate()
if zoo_nodes != []:
print "Terminating zoo..."
for inst in zoo_nodes:
inst.terminate()
# Delete security groups as well
if opts.delete_groups:
print "Deleting security groups (this will take some time)..."
group_names = [cluster_name + "-master", cluster_name + "-slaves", cluster_name + "-zoo"]
attempt = 1;
while attempt <= 3:
print "Attempt %d" % attempt
groups = [g for g in conn.get_all_security_groups() if g.name in group_names]
success = True
# Delete individual rules in all groups before deleting groups to
# remove dependencies between them
for group in groups:
print "Deleting rules in security group " + group.name
for rule in group.rules:
for grant in rule.grants:
success &= group.revoke(ip_protocol=rule.ip_protocol,
from_port=rule.from_port,
to_port=rule.to_port,
src_group=grant)
# Sleep for AWS eventual-consistency to catch up, and for instances
# to terminate
time.sleep(30) # Yes, it does have to be this long :-(
for group in groups:
try:
conn.delete_security_group(group.name)
print "Deleted security group " + group.name
except boto.exception.EC2ResponseError:
success = False;
print "Failed to delete security group " + group.name
# Unfortunately, group.revoke() returns True even if a rule was not
# deleted, so this needs to be rerun if something fails
if success: break;
attempt += 1
if not success:
print "Failed to delete all security groups after 3 tries."
print "Try re-running in a few minutes."
elif action == "login":
(master_nodes, slave_nodes, zoo_nodes) = get_existing_cluster(
conn, opts, cluster_name)
master = master_nodes[0].public_dns_name
print "Logging into master " + master + "..."
proxy_opt = ""
if opts.proxy_port != None:
proxy_opt = "-D " + opts.proxy_port
subprocess.check_call("ssh -o StrictHostKeyChecking=no -i %s %s %s@%s" %
(opts.identity_file, proxy_opt, opts.user, master), shell=True)
elif action == "get-master":
(master_nodes, slave_nodes, zoo_nodes) = get_existing_cluster(conn, opts, cluster_name)
print master_nodes[0].public_dns_name
elif action == "stop":
response = raw_input("Are you sure you want to stop the cluster " +
cluster_name + "?\nDATA ON EPHEMERAL DISKS WILL BE LOST, " +
"BUT THE CLUSTER WILL KEEP USING SPACE ON\n" +
"AMAZON EBS IF IT IS EBS-BACKED!!\n" +
"Stop cluster " + cluster_name + " (y/N): ")
if response == "y":
(master_nodes, slave_nodes, zoo_nodes) = get_existing_cluster(
conn, opts, cluster_name, die_on_error=False)
print "Stopping master..."
for inst in master_nodes:
if inst.state not in ["shutting-down", "terminated"]:
inst.stop()
print "Stopping slaves..."
for inst in slave_nodes:
if inst.state not in ["shutting-down", "terminated"]:
inst.stop()
if zoo_nodes != []:
print "Stopping zoo..."
for inst in zoo_nodes:
if inst.state not in ["shutting-down", "terminated"]:
inst.stop()
elif action == "start":
(master_nodes, slave_nodes, zoo_nodes) = get_existing_cluster(
conn, opts, cluster_name)
print "Starting slaves..."
for inst in slave_nodes:
if inst.state not in ["shutting-down", "terminated"]:
inst.start()
print "Starting master..."
for inst in master_nodes:
if inst.state not in ["shutting-down", "terminated"]:
inst.start()
if zoo_nodes != []:
print "Starting zoo..."
for inst in zoo_nodes:
if inst.state not in ["shutting-down", "terminated"]:
inst.start()
wait_for_cluster(conn, opts.wait, master_nodes, slave_nodes, zoo_nodes)
setup_cluster(conn, master_nodes, slave_nodes, zoo_nodes, opts, False)
else:
print >> stderr, "Invalid action: %s" % action
sys.exit(1)
if __name__ == "__main__":
logging.basicConfig()
main()
|