def __init__(
self,
aws,
wait_time_secs,
app_name=None,
):
"""
Creates the EbsHelper
"""
self.aws = aws
self.ebs = connect_to_region(aws.region,
aws_access_key_id=aws.access_key,
aws_secret_access_key=aws.secret_key,
security_token=aws.security_token)
self.autoscale = AutoScaleConnection(
aws_access_key_id=aws.access_key,
aws_secret_access_key=aws.secret_key,
security_token=aws.security_token)
self.s3 = S3Connection(
aws_access_key_id=aws.access_key,
aws_secret_access_key=aws.secret_key,
security_token=aws.security_token,
host=(lambda r: 's3.amazonaws.com'
if r == 'us-east-1' else 's3-' + r + '.amazonaws.com')(
aws.region))
self.app_name = app_name
self.wait_time_secs = wait_time_secs
def __init__(self, clc_host, access_id, secret_key, token):
#boto.set_stream_logger('foo')
path = '/services/AutoScaling'
port = 8773
if clc_host[len(clc_host) - 13:] == 'amazonaws.com':
clc_host = clc_host.replace('ec2', 'autoscaling', 1)
path = '/'
reg = None
port = 443
reg = RegionInfo(name='eucalyptus', endpoint=clc_host)
if boto.__version__ < '2.6':
self.conn = AutoScaleConnection(access_id,
secret_key,
region=reg,
port=port,
path=path,
is_secure=True,
security_token=token,
debug=0)
else:
self.conn = AutoScaleConnection(access_id,
secret_key,
region=reg,
port=port,
path=path,
validate_certs=False,
is_secure=True,
security_token=token,
debug=0)
self.conn.http_connection_kwargs['timeout'] = 30
def scale_instances(tasks_per_instance, group_name, total_groups):
conn = AutoScaleConnection()
group = conn.get_all_groups(names=[group_name])[0]
if group.desired_capacity == group.max_size:
logger.info('Maximum number of instances reached')
return
tasks_count = get_sqs_tasks_count()
if not tasks_count:
logger.info('No tasks left in queues')
return
logger.info('Num of tasks in queues %s', tasks_count)
tasks_per_instance = float(tasks_per_instance)
additional_instances_count = int(
ceil(tasks_count / tasks_per_instance) / total_groups)
updated_instances_count = \
group.desired_capacity + additional_instances_count
# consider max allowed instances
if updated_instances_count > group.max_size:
updated_instances_count = group.max_size
logger.info('Updating group from %s to %s instances',
group.desired_capacity, updated_instances_count)
group.set_capacity(updated_instances_count)
group.desired_capacity = updated_instances_count
group.update()
logger.info('Done\n')
def connect(self, groupname): self.ec2 = boto.connect_ec2() self.cw = CloudWatchConnection() self.autoscale = AutoScaleConnection() self.group = self.autoscale.get_all_groups(names=[groupname])[0] self.instances = len(self.group.instances) self.desired = self.group.desired_capacity self.name = groupname
def test_basic(self):
# NB: as it says on the tin these are really basic tests that only
# (lightly) exercise read-only behaviour - and that's only if you
# have any autoscale groups to introspect. It's useful, however, to
# catch simple errors
print '--- running %s tests ---' % self.__class__.__name__
c = AutoScaleConnection()
self.assertTrue(repr(c).startswith('AutoScaleConnection'))
groups = c.get_all_groups()
for group in groups:
self.assertTrue(type(group), AutoScalingGroup)
# get activities
activities = group.get_activities()
for activity in activities:
self.assertEqual(type(activity), Activity)
# get launch configs
configs = c.get_all_launch_configurations()
for config in configs:
self.assertTrue(type(config), LaunchConfiguration)
# get policies
policies = c.get_all_policies()
for policy in policies:
self.assertTrue(type(policy), ScalingPolicy)
# get scheduled actions
actions = c.get_all_scheduled_actions()
for action in actions:
self.assertTrue(type(action), ScheduledUpdateGroupAction)
# get instances
instances = c.get_all_autoscaling_instances()
for instance in instances:
self.assertTrue(type(instance), Instance)
# get all scaling process types
ptypes = c.get_all_scaling_process_types()
for ptype in ptypes:
self.assertTrue(type(ptype), ProcessType)
# get adjustment types
adjustments = c.get_all_adjustment_types()
for adjustment in adjustments:
self.assertTrue(type(adjustment), AdjustmentType)
# get metrics collection types
types = c.get_all_metric_collection_types()
self.assertTrue(type(types), MetricCollectionTypes)
print '--- tests completed ---'
def test_ebs_optimized_regression(self):
c = AutoScaleConnection()
time_string = '%d' % int(time.time())
lc_name = 'lc-%s' % time_string
lc = LaunchConfiguration(name=lc_name,
image_id='ami-2272864b',
instance_type='t1.micro',
ebs_optimized=True)
# This failed due to the difference between native Python ``True/False``
# & the expected string variants.
c.create_launch_configuration(lc)
self.addCleanup(c.delete_launch_configuration, lc_name)
def cleanup_unused_launch_configs(unused_launch_config_names, delete=False):
conn = AutoScaleConnection()
configs = conn.get_all_launch_configurations(names=unused_launch_config_names)
print "\nGetting ready to cleanup launch configs ... {}".format(delete and "FOR REAL" or "DRYRUN")
for config in configs:
if delete:
print "deleting launch config: {} in {} seconds...".format(config.name, 5)
time.sleep(5)
print "deleting launch config: {}!".format(config.name)
response = config.delete()
print "deleted launch config: {} ({})!".format(config.name, response)
else:
print "dry run: not deleting config:", config.name
def test_ebs_optimized_regression(self):
c = AutoScaleConnection()
time_string = '%d' % int(time.time())
lc_name = 'lc-%s' % time_string
lc = LaunchConfiguration(
name=lc_name,
image_id='ami-2272864b',
instance_type='t1.micro',
ebs_optimized=True
)
# This failed due to the difference between native Python ``True/False``
# & the expected string variants.
c.create_launch_configuration(lc)
self.addCleanup(c.delete_launch_configuration, lc_name)
def get_all_group_instances_and_conn():
conn = AutoScaleConnection()
global autoscale_conn
autoscale_conn = conn
ec2 = boto.ec2.connect_to_region('us-east-1')
groups = conn.get_all_groups(
names=['SCCluster1', 'SCCluster2', 'SCCluster3',
'SCCluster4']) # TODO: update this list
instances = [instance for group in groups for instance in group]
if not instances:
sys.exit()
instance_ids = [instance.instance_id for instance in instances]
instances = ec2.get_only_instances(instance_ids)
return instances, conn
def _remove_from_worker_pool():
"""
Ensures that this instance is shut down, and unregisted from the worker
pool.
"""
# Retrieve the current state of the pool.
pool = AutoScaleConnection().get_all_groups(["LSDA Worker Pool"])[0]
if pool.desired_capacity <= pool.min_size:
return
# Reduce the pool size and shut ourself down.
pool.desired_capacity -= 1
pool.update()
def as_connect(region=None, *args, **kwargs):
"""Helper to connect to Amazon Web Services EC2, using identify provided
by environment, as also optional region in arguments.
"""
if not os_environ.get("AWS_ACCESS_KEY_ID", None):
raise EC2LibraryError(
"Environment variable AWS_ACCESS_KEY_ID is not set.")
if not os_environ.get("AWS_SECRET_ACCESS_KEY", None):
raise EC2LibraryError(
"Environment variable AWS_SECRET_ACCESS_KEY is not set.")
if not region:
region = env.get("ec2_region")
for reg in boto.ec2.autoscale.regions():
if reg.name == region:
region = reg
connection = AutoScaleConnection(os_environ.get("AWS_ACCESS_KEY_ID"),
os_environ.get("AWS_SECRET_ACCESS_KEY"),
region=region,
*args,
**kwargs)
return connection
def main():
parser = optparse.OptionParser()
parser.add_option( "-c", "--config", dest="config_file", help="AutoScale config INI", metavar="FILE" )
(options, args) = parser.parse_args()
logging.info( "Using config file [%s]" % options.config_file )
config = parse_config( options.config_file )
aws_access = config.get("AWS", 'access')
aws_secret = config.get("AWS", 'secret')
logging.debug( "Connecting to AWS with access [%s] and secret [%s]" % ( aws_access, aws_secret ) )
aws_connection = AutoScaleConnection( aws_access, aws_secret )
print "AutoScalingGroups:"
print aws_connection.get_all_groups().__dict__
def autoscale_group_hosts(group_name):
import boto.ec2
from boto.ec2.autoscale import AutoScaleConnection
ec2 = boto.connect_ec2()
conn = AutoScaleConnection()
groups = conn.get_all_groups(names=[])
groups = [ group for group in groups if group.name.startswith(group_name) ]
instance_ids = []
instances = []
for group in groups:
print group.name
instance_ids.extend([i.instance_id for i in group.instances])
instances.extend(ec2.get_only_instances(instance_ids))
return [i.private_ip_address for i in instances], instances[0].id, instances[0].tags.get("aws:autoscaling:groupName")
def __init__(self,PREFIX='tfound-',ENV='dev',AMI='',TYPE='',SIZE='',
DOMAIN='tfound',SSHKEY='myprivatekey',AWSKEY='',AWSSECRET='',AVAIL_ZONES=["us-east-1a","us-east-1b","us-east-1c","us-east-1d"]):
'''
Shows examples
Create load balancer group 'tfound-dev-web-lb' for web servers, in dev group for tfound:
python control-lb-and-groups.py --createlb --env dev --aws SC --type web
Add an instance to the load balancer group:
python control-lb-and-groups.py --addtolb=true --env dev --aws SC --type web --instance=i-999999
Create launch config using ami ami-fa6b8393 (default), medium sized instance, and Autoscale Group 'tfound-dev-web-group' with a min of 2 instances, max 5, with health check on port 80:
python control-lb-and-groups.py --createlc --ami ami-fa6b8393 --size c1.medium --env dev --aws SC --type web --createag --min 2 --max 5
Triggers/Health checks are hard coded to spawn new instances when total cpu reaches 60 percent or health check fails.
'''
self.PREFIX=PREFIX+DOMAIN+'-'+ENV+'-'+TYPE
self.ENV=ENV
self.AMI=AMI
self.TYPE=TYPE
self.DOMAIN=DOMAIN
self.SIZE=SIZE
self.MIN=MIN
self.MAX=MAX
self.SSHKEY=SSHKEY
self.AWSKEY=AWSKEY
self.AWSSECRET=AWSSECRET
self.AVAIL_ZONES=AVAIL_ZONES
self.LBNAME=self.PREFIX+'-lb'
self.AGNAME=self.PREFIX+'-group'
self.TRNAME=self.PREFIX+'-trigger'
self.LCNAME=self.PREFIX+'-launch_config'
self.asconn=AutoScaleConnection(self.AWSKEY, self.AWSSECRET)
self.elbconn = ELBConnection(aws_access_key_id=AWSKEY,aws_secret_access_key=AWSSECRET)
self.lc = self._buildLaunchConfig()
self.ag = self._buildAutoscaleGroup()
def main():
"""
Main entry point for the automated scaling daemon.
"""
# Configure logging.
logging.basicConfig(
format = "%(asctime)-15s %(levelname)5s %(message)s",
level = logging.INFO
)
# Read configuration.
options = yaml.load(open("config.yaml"))
# Connect to the RabbitMQ cluster.
params = pika.ConnectionParameters(host=options["amqp"])
conn = pika.BlockingConnection(params)
channel = conn.channel()
while True:
# Ensure that we have things stuck in the queue for the given amount
# of time.
for i in xrange(DELAY / 5):
queue_length = get_queue_length(channel, "stable")
logging.info("Queue length: {}".format(queue_length))
if queue_length == 0:
break
time.sleep(5)
else:
# Scale up!
group = AutoScaleConnection().get_all_groups(["LSDA Worker Pool"])[0]
group.desired_capacity = min(
group.desired_capacity + 2, group.max_size)
group.update()
logging.info(
"Triggering increase to {}".format(group.desired_capacity))
time.sleep(300)
# Wait until next polling event.
time.sleep(30)
def _is_up_to_date():
"""
Returns True if this instance is up to date.
"""
# Retrieve instance information.
conn = AutoScaleConnection()
pool = conn.get_all_groups(["LSDA Worker Pool"])[0]
config = conn.get_all_launch_configurations(
names=[pool.launch_config_name])[0]
# Retrive the AMI for this instance and for others.
config_ami = config.image_id
my_ami = urllib.urlopen("http://169.254.169.254/latest/"
"meta-data/ami-id").read()
return config_ami == my_ami
def get_all_group_instances_and_conn(
groups_names=get_autoscale_groups()['groups']):
conn = AutoScaleConnection()
global autoscale_conn
autoscale_conn = conn
ec2 = boto.ec2.connect_to_region('us-east-1')
selected_group_name = random.choice(groups_names)
logger.info('Selected autoscale group: %s' % selected_group_name)
group = conn.get_all_groups(names=[selected_group_name])[0]
if not group.instances:
logger.info("No working instances in selected group %s" %
selected_group_name)
upload_logs_to_s3()
sys.exit()
instance_ids = [i.instance_id for i in group.instances]
instances = ec2.get_only_instances(instance_ids)
return instances, conn
def autoscale_group_hosts(group_name):
import boto.ec2
from boto.ec2.autoscale import AutoScaleConnection
ec2 = boto.connect_ec2()
conn = AutoScaleConnection()
groups = conn.get_all_groups(names=[])
groups = [group for group in groups if group.name.startswith(group_name)]
instance_ids = []
instances = []
for group in groups:
print "group name:", group.name
instance_ids.extend([i.instance_id for i in group.instances])
instances.extend(ec2.get_only_instances(instance_ids))
return set([
i.private_ip_address for i in instances
]), instances[0].id, instances[0].tags.get("aws:autoscaling:groupName")
def as_connection():
""" Create and return an Auto Scale Connection """
key_id = config.get('Credentials', 'aws_access_key_id')
access_key = config.get('Credentials', 'aws_secret_access_key')
conn = AutoScaleConnection(key_id, access_key)
return conn
def __init__(self, clc_host, access_id, secret_key, token):
#boto.set_stream_logger('foo')
path='/services/AutoScaling'
port=8773
if clc_host[len(clc_host)-13:] == 'amazonaws.com':
clc_host = clc_host.replace('ec2', 'autoscaling', 1)
path = '/'
reg = None
port=443
reg = RegionInfo(name='eucalyptus', endpoint=clc_host)
if boto.__version__ < '2.6':
self.conn = AutoScaleConnection(access_id, secret_key, region=reg,
port=port, path=path,
is_secure=True, security_token=token, debug=0)
else:
self.conn = AutoScaleConnection(access_id, secret_key, region=reg,
port=port, path=path, validate_certs=False,
is_secure=True, security_token=token, debug=0)
self.conn.http_connection_kwargs['timeout'] = 30
def __init__(self, args):
"""
Initializing basic variables needed for auto scaling
"""
self.configs = ConfigParser.RawConfigParser()
self.args = args
self.test_props = {}
self.props = {}
self.ec2_connection = EC2Connection(self.args.access_key,
self.args.secret_key)
self.autoscale_connection = AutoScaleConnection(
self.args.access_key, self.args.secret_key)
self.elb_connection = ELBConnection(self.args.access_key,
self.args.secret_key)
self.cw_connection = CloudWatchConnection(self.args.access_key,
self.args.secret_key)
self.firstInstance = None
self.launchConfiguration = None
self.healthCheck = None
def find_unused_launch_configs():
conn = AutoScaleConnection()
autoscale_groups = conn.get_all_groups(max_records=100)
launch_configs = conn.get_all_launch_configurations(max_records=100)
launch_config_names = {lc.name for lc in launch_configs}
used_launch_config_names = {asg.launch_config_name for asg in autoscale_groups}
unused_launch_config_names = launch_config_names - used_launch_config_names
print "Autoscale Groups and Current Launch Configs:"
print "{:<40}{:<40}".format("ASG", "LC")
for asg in autoscale_groups:
#print "asg:", asg.name, "-> lc:", asg.launch_config_name
print "{:<40}{:<40}".format(asg.name, asg.launch_config_name)
print "\nUnused Launch Configs: (launch configs without a autoscale group)"
unused_launch_config_names = list(sorted(unused_launch_config_names))
for unused_launch_config in unused_launch_config_names:
print "\t", unused_launch_config
return unused_launch_config_names
def _create_autoscale_connection(self):
LOG.debug("Creating autoscale connection for %s" % self.config.name)
region = RegionInfo(name=self.config.cloud_type,
endpoint=self.config.as_uri)
self._as_conn = AutoScaleConnection(
aws_access_key_id=self.config.access_id,
aws_secret_access_key=self.config.secret_key,
is_secure=True,
port=self.config.as_port,
region=region,
validate_certs=False)
def connect_autoscale(aws_access_key_id=None, aws_secret_access_key=None, **kwargs):
"""
:type aws_access_key_id: string
:param aws_access_key_id: Your AWS Access Key ID
:type aws_secret_access_key: string
:param aws_secret_access_key: Your AWS Secret Access Key
:rtype: :class:`boto.ec2.autoscale.AutoScaleConnection`
:return: A connection to Amazon's Auto Scaling Service
"""
from boto.ec2.autoscale import AutoScaleConnection
return AutoScaleConnection(aws_access_key_id, aws_secret_access_key, **kwargs)
def main():
parser = optparse.OptionParser()
parser.add_option("-c", "--config", dest="config_file", help="AutoScale config INI", metavar="FILE")
(options, args) = parser.parse_args()
logging.info("Using config file [%s]" % options.config_file)
config = parse_config(options.config_file)
aws_access = config.get("AWS", "access")
aws_secret = config.get("AWS", "secret")
logging.debug("Connecting to AWS with access [%s] and secret [%s]" % (aws_access, aws_secret))
aws_connection = AutoScaleConnection(aws_access, aws_secret)
lc = boto.ec2.autoscale.launchconfig.LaunchConfiguration(
name=config.get("LaunchConfig", "name"),
image_id=config.get("LaunchConfig", "image"),
key_name=config.get("LaunchConfig", "key"),
user_data=config.get("LaunchConfig", "user_data"),
security_groups=config.get("LaunchConfig", "security_groups"),
instance_type=config.get("LaunchConfig", "instance_type"),
)
logging.info("LC CONFIG = %s" % lc.__dict__)
asg = boto.ec2.autoscale.group.AutoScalingGroup(
group_name=config.get("AutoScaleGroup", "group_name"),
availability_zones=config.get("AutoScaleGroup", "zones"),
min_size=config.get("AutoScaleGroup", "min_instances"),
max_size=config.get("AutoScaleGroup", "max_instances"),
launch_config=lc,
)
print "ASG dict: %s" % asg.__dict__
asg.connection = aws_connection
params = {"AutoScalingGroupName": asg.name}
asg = aws_connection.get_object("DescribeAutoScalingGroups", params, boto.ec2.autoscale.group.AutoScalingGroup)
print asg
def main():
# sys.exit() # disable self-killer AGAIN
conn = AutoScaleConnection()
instance_id = get_instance_metadata()['instance-id']
log_file = '/tmp/remote_instance_starter2.log'
flag, reason = check_logs_status(log_file)
if flag and reason:
s3_conn = boto.connect_s3()
bucket = s3_conn.get_bucket(BUCKET_NAME)
k = Key(bucket)
k.key = BUCKET_KEY
global log_file_path
time.sleep(70)
# Try to upload logs prior to stop server
os.system('python upload_logs_to_s3.py')
os.system('rm %s' % log_file)
k.get_contents_to_filename(log_file_path)
logger.warning(
"Instance with id=%s was terminated"
" due to reason='%s'. "
"Instance was killed by itself.", instance_id, reason)
k.set_contents_from_filename(log_file_path)
conn.terminate_instance(instance_id, decrement_capacity=True)
def __init__(self, clc_host, access_id, secret_key, token):
#boto.set_stream_logger('foo')
path = '/services/AutoScaling'
reg = RegionInfo(name='eucalyptus', endpoint=clc_host)
port = 8773
if clc_host[len(clc_host) - 13:] == 'amazonaws.com':
clc_host = clc_host.replace('ec2', 'autoscaling', 1)
path = '/'
reg = None
port = 443
self.conn = AutoScaleConnection(access_id,
secret_key,
region=reg,
port=port,
path=path,
is_secure=True,
security_token=token,
debug=0)
self.conn.APIVersion = '2011-01-01'
if not (clc_host[len(clc_host) - 13:] == 'amazonaws.com'):
self.conn.auth_region_name = 'Eucalyptus'
self.conn.https_validate_certificates = False
self.conn.http_connection_kwargs['timeout'] = 30
def __init__(self, args):
"""
Initializing basic variables needed for auto scaling
"""
self.configs = ConfigParser.RawConfigParser()
self.args = args
self.test_props = {}
self.props = {}
self.ec2_connection = EC2Connection(self.args.access_key, self.args.secret_key)
self.autoscale_connection = AutoScaleConnection(self.args.access_key, self.args.secret_key)
self.elb_connection = ELBConnection(self.args.access_key, self.args.secret_key)
self.cw_connection = CloudWatchConnection(self.args.access_key, self.args.secret_key)
self.firstInstance = None
self.launchConfiguration = None
self.healthCheck = None
def launch_auto_scaling(stage = 'development'):
config = get_provider_dict()
from boto.ec2.autoscale import AutoScaleConnection, AutoScalingGroup, LaunchConfiguration, Trigger
conn = AutoScaleConnection(fabric.api.env.conf['AWS_ACCESS_KEY_ID'], fabric.api.env.conf['AWS_SECRET_ACCESS_KEY'], host='%s.autoscaling.amazonaws.com' % config['location'][:-1])
for name, values in config.get(stage, {}).get('autoscale', {}):
if any(group.name == name for group in conn.get_all_groups()):
fabric.api.warn(fabric.colors.orange('Autoscale group %s already exists' % name))
continue
lc = LaunchConfiguration(name = '%s-launch-config' % name, image_id = values['image'], key_name = config['key'])
conn.create_launch_configuration(lc)
ag = AutoScalingGroup(group_name = name, load_balancers = values.get('load-balancers'), availability_zones = [config['location']], launch_config = lc, min_size = values['min-size'], max_size = values['max-size'])
conn.create_auto_scaling_group(ag)
if 'min-cpu' in values and 'max-cpu' in values:
tr = Trigger(name = '%s-trigger' % name, autoscale_group = ag, measure_name = 'CPUUtilization', statistic = 'Average', unit = 'Percent', dimensions = [('AutoScalingGroupName', ag.name)],
period = 60, lower_threshold = values['min-cpu'], lower_breach_scale_increment = '-1', upper_threshold = values['max-cpu'], upper_breach_scale_increment = '2', breach_duration = 60)
conn.create_trigger(tr)
def connect_autoscale(aws_access_key_id=None,
aws_secret_access_key=None,
**kwargs):
"""
@type aws_access_key_id: string
@param aws_access_key_id: Your AWS Access Key ID
@type aws_secret_access_key: string
@param aws_secret_access_key: Your AWS Secret Access Key
@rtype: L{AutoScaleConnnection<boto.ec2.autoscale.AutoScaleConnection>}
@return: A connection to Amazon's Auto Scaling Service
"""
from boto.ec2.autoscale import AutoScaleConnection
return AutoScaleConnection(aws_access_key_id, aws_secret_access_key,
**kwargs)
def launch_auto_scaling(stage='development'):
config = get_provider_dict()
from boto.ec2.autoscale import AutoScaleConnection, AutoScalingGroup, LaunchConfiguration, Trigger
conn = AutoScaleConnection(fabric.api.env.conf['AWS_ACCESS_KEY_ID'],
fabric.api.env.conf['AWS_SECRET_ACCESS_KEY'],
host='%s.autoscaling.amazonaws.com' %
config['location'][:-1])
for name, values in config.get(stage, {}).get('autoscale', {}):
if any(group.name == name for group in conn.get_all_groups()):
fabric.api.warn(
fabric.colors.orange('Autoscale group %s already exists' %
name))
continue
lc = LaunchConfiguration(name='%s-launch-config' % name,
image_id=values['image'],
key_name=config['key'])
conn.create_launch_configuration(lc)
ag = AutoScalingGroup(group_name=name,
load_balancers=values.get('load-balancers'),
availability_zones=[config['location']],
launch_config=lc,
min_size=values['min-size'],
max_size=values['max-size'])
conn.create_auto_scaling_group(ag)
if 'min-cpu' in values and 'max-cpu' in values:
tr = Trigger(name='%s-trigger' % name,
autoscale_group=ag,
measure_name='CPUUtilization',
statistic='Average',
unit='Percent',
dimensions=[('AutoScalingGroupName', ag.name)],
period=60,
lower_threshold=values['min-cpu'],
lower_breach_scale_increment='-1',
upper_threshold=values['max-cpu'],
upper_breach_scale_increment='2',
breach_duration=60)
conn.create_trigger(tr)
def delete_autoscaling():
con = AutoScaleConnection(aws_secret_access_key=AWS_SECRET_ACCESS_KEY,
aws_access_key_id=AWS_ACCESS_KEY,
region=RegionInfo(name=REGION,
endpoint='autoscaling.%s.amazonaws.com' % REGION))
print "Deleting autoscaling group.."
group = con.get_all_groups(names=[AUTOSCALING_GROUP_NAME])[0]
print "shutting down instances"
group.shutdown_instances()
time.sleep(LONG_SLEEP_PERIOD)
print "Deleting autoscaling group itself"
con.delete_auto_scaling_group(AUTOSCALING_GROUP_NAME, force_delete=True)
print "Deleting launch configuration"
con.delete_launch_configuration(AUTOSCALING_GROUP_NAME)
con.close()
def __init__(self, aws, wait_time_secs, app_name=None,):
"""
Creates the EbsHelper
"""
self.aws = aws
self.ebs = connect_to_region(aws.region, aws_access_key_id=aws.access_key,
aws_secret_access_key=aws.secret_key,
security_token=aws.security_token)
self.autoscale = AutoScaleConnection(aws_access_key_id=aws.access_key,
aws_secret_access_key=aws.secret_key,
security_token=aws.security_token)
self.s3 = S3Connection(
aws_access_key_id=aws.access_key,
aws_secret_access_key=aws.secret_key,
security_token=aws.security_token,
host=(lambda r: 's3.amazonaws.com' if r == 'us-east-1' else 's3-' + r + '.amazonaws.com')(aws.region))
self.app_name = app_name
self.wait_time_secs = wait_time_secs
def set_endpoint(self, endpoint):
#boto.set_stream_logger('scale')
path = '/services/AutoScaling'
reg = RegionInfo(name='eucalyptus', endpoint=endpoint)
port = 8773
if endpoint[len(endpoint)-13:] == 'amazonaws.com':
endpoint = endpoint.replace('ec2', 'autoscaling', 1)
path = '/'
reg = RegionInfo(endpoint=endpoint)
port = 443
self.conn = AutoScaleConnection(self.access_id, self.secret_key, region=reg,
port=port, path=path,
is_secure=True, security_token=self.token, debug=0)
self.conn.APIVersion = '2011-01-01'
if not(endpoint[len(endpoint)-13:] == 'amazonaws.com'):
self.conn.auth_region_name = 'Eucalyptus'
self.conn.https_validate_certificates = False
self.conn.http_connection_kwargs['timeout'] = 30
def create_autoscaling_group():
global img
conn = AutoScaleConnection(os.environ['AWS_ACCESS_KEY_ID'], os.environ['AWS_SECRET_ACCESS_KEY'])
autoscale = boto.ec2.autoscale.connect_to_region('us-east-1')
print conn.get_all_groups()
timestamp = time.time()
value = datetime.datetime.fromtimestamp(timestamp)
humanreadabledate = value.strftime('%Y-%m-%d_%H.%M.%S')
config_name = 'live_launch_config'+humanreadabledate
init_script = "#!/bin/sh /home/ec2-user/sds/deployment_scripts/initialize_server.py"
lc = LaunchConfiguration(name=config_name, image_id=img,
key_name='SDSEastKey',
security_groups=['sg-a7afb1c2'],
user_data=init_script)
conn.create_launch_configuration(lc)
ag = AutoScalingGroup(group_name=config_name, load_balancers=['SDSLiveLoadBalancer'], availability_zones=['us-east-1a'], launch_config=lc, min_size=2, max_size=2, connection=conn)
conn.create_auto_scaling_group(ag)
def connect_autoscale(aws_access_key_id=None, aws_secret_access_key=None,
**kwargs):
"""
:type aws_access_key_id: string
:param aws_access_key_id: Your AWS Access Key ID
:type aws_secret_access_key: string
:param aws_secret_access_key: Your AWS Secret Access Key
:rtype: :class:`boto.ec2.autoscale.AutoScaleConnection`
:return: A connection to Amazon's Auto Scaling Service
:type use_block_device_types bool
:param use_block_device_types: Specifies whether to return described Launch Configs with block device mappings containing
block device types, or a list of old style block device mappings (deprecated). This defaults to false for compatability
with the old incorrect style.
"""
from boto.ec2.autoscale import AutoScaleConnection
return AutoScaleConnection(aws_access_key_id, aws_secret_access_key,
**kwargs)
def get_asg_connection():
conn = AutoScaleConnection()
autoscale_groups = conn.get_all_groups(max_records=1)
return conn
}
######################### end parameter block ################################
######################### begin configuration ################################
# make the connections
conn_ec2 = boto.ec2.connect_to_region(
regionName,
aws_access_key_id = AWS_ACCESS_KEY,
aws_secret_access_key = AWS_SECRET_KEY
)
conn_reg = boto.ec2.elb.connect_to_region(regionName)
conn_elb = ELBConnection(AWS_ACCESS_KEY, AWS_SECRET_KEY)
conn_as = AutoScaleConnection(AWS_ACCESS_KEY, AWS_SECRET_KEY)
conn_cw = boto.ec2.cloudwatch.connect_to_region(regionName)
conn_cw = boto.ec2.cloudwatch.connect_to_region(
regionName,
aws_access_key_id = AWS_ACCESS_KEY,
aws_secret_access_key = AWS_SECRET_KEY
)
######################### end configuration ################################
# balancers = elb.get_all_load_balancers()
# print balancers[0]
# retrieve the instances in the autoscale group
def create_AutoScaling():
print "Creating AutoScaling..."
# establish connection
as_conn = AutoScaleConnection(AWSAccessKeyId, AWSSecretKey)
# create launch configuration
global lc
lc = LaunchConfiguration(name='lc',
image_id=DATA_CEN_AMI,
key_name=ACCESS_KEY,
instance_monitoring=True,
security_groups=[SECURITY_GRP],
instance_type=MACHINE_TYPE)
as_conn.create_launch_configuration(lc)
# create tag for autoscaling group
as_tag = Tag(key="Project",
value="2.2",
propagate_at_launch=True,
resource_id='my_group')
# create aotoscaling group
global ag
ag = AutoScalingGroup(group_name='my_group',
load_balancers=['myELB'],
availability_zones=['us-east-1a'],
launch_config=lc,
min_size=MIN_SIZE,
max_size=MAX_SIZE,
connection=as_conn,
tags=[as_tag])
# associate the autoscaling group with launch configuration
as_conn.create_auto_scaling_group(ag)
# build the scale policy
scale_up_policy = ScalingPolicy(name='scale_up',
adjustment_type='ChangeInCapacity',
as_name='my_group',
scaling_adjustment=1,
cooldown=60)
scale_down_policy = ScalingPolicy(name='scale_down',
adjustment_type='ChangeInCapacity',
as_name='my_group',
scaling_adjustment=-1,
cooldown=60)
# register the scale policy
as_conn.create_scaling_policy(scale_up_policy)
as_conn.create_scaling_policy(scale_down_policy)
# refresh the scale policy for extra information
scale_up_policy = as_conn.get_all_policies(as_group='my_group',
policy_names=['scale_up'])[0]
scale_down_policy = as_conn.get_all_policies(as_group='my_group',
policy_names=['scale_down'
])[0]
# create cloudwatch alarm
cloudwatch = CloudWatchConnection(aws_access_key_id=AWSAccessKeyId,
aws_secret_access_key=AWSSecretKey,
is_secure=True)
# region='us-east-1a')
# assocate cloudwatch with alarm
alarm_dimensions = {"AutoScalingGroupName": 'my_group'}
# create scale up alarm
scale_up_alarm = MetricAlarm(name='scale_up_on_cpu',
namespace='AWS/EC2',
metric='CPUUtilization',
statistic='Average',
comparison='>',
threshold='50',
period='60',
evaluation_periods=2,
alarm_actions=[scale_up_policy.policy_arn],
dimensions=alarm_dimensions)
cloudwatch.create_alarm(scale_up_alarm)
# create scale down alarm
scale_down_alarm = MetricAlarm(
name='scale_down_on_cpu',
namespace='AWS/EC2',
metric='CPUUtilization',
statistic='Average',
comparison='<',
threshold='20',
period='60',
evaluation_periods=1,
alarm_actions=[scale_down_policy.policy_arn],
dimensions=alarm_dimensions)
cloudwatch.create_alarm(scale_down_alarm)
print "AutoScaling created successfully"
def main():
parser = argparse.ArgumentParser(
description = "triggers a full LSDA rollout")
parser.add_argument("--inspect", action = "store_true",
help = "pause before baking AMI", default = False)
parser.add_argument("--clean", action = "store_true",
help = "reset from clean Ubuntu 12.04 image", default = False)
parser.add_argument("--no-restart", action = "store_true",
dest = "no_restart", help = "don't restart all nodes in ASG",
default = False)
options = parser.parse_args()
logging.info("Starting rollout.")
conn_ec2 = boto.ec2.connect_to_region("us-east-1")
conn_ec2_as = AutoScaleConnection()
if not options.clean:
logging.info("Searching for existing images...")
group = conn_ec2_as.get_all_groups(['LSDA Worker Pool'])[0]
launch_config = conn_ec2_as.get_all_launch_configurations(
names=[group.launch_config_name])[0]
existing_images = conn_ec2.get_all_images(owners = ["self"])[0]
ami_id = launch_config.image_id
logging.info("Using existing image {0}".format(ami_id))
else:
ami_id = 'ami-59a4a230' # Clean Ubuntu 12.04.
logging.info("Using base image {0}".format(ami_id))
reservation = conn_ec2.run_instances(
image_id = ami_id,
key_name = 'jeremy-aws-key',
instance_type = 't1.micro',
security_groups = ['Worker Nodes'],
)
try:
instance = reservation.instances[0]
logging.info("Waiting for instance {} to start...".format(instance.id))
instance.update()
while instance.ip_address is None:
logging.info("Not ready. Retrying in 10 seconds...")
time.sleep(10)
instance.update()
while True:
result = subprocess.call(["ssh", "-o",
"UserKnownHostsFile=/dev/null", "-o", "StrictHostKeyChecking=no",
"ubuntu@{}".format(instance.ip_address), "uname -r"])
if result != 0:
logging.info("Not ready for SSH. Retrying in 10 seconds...")
time.sleep(10)
else:
break
logging.info("Instance has started; running setup script.")
logging.info("(IP address is {})".format(instance.ip_address))
subprocess.check_call(["ssh", "-o", "UserKnownHostsFile=/dev/null",
"-o", "StrictHostKeyChecking=no",
"ubuntu@{}".format(instance.ip_address),
"sudo stop lsda; sleep 20; sudo rm worker.sh;"
"wget https://raw.github.com/fatlotus/lsda-infrastructure/"
"master/servers/worker.sh; sudo bash worker.sh"])
if options.inspect:
logging.info("Connect to ubuntu@{} to inspect the image."
.format(instance.ip_address))
logging.info("When you're done, press CTRL-C.")
try:
while True:
time.sleep(3600)
except KeyboardInterrupt:
pass
logging.info("Creating AMI from existing image.")
new_image = instance.create_image(
name = ('Latest-{:%Y-%m-%d--%H-%M-%S}'.
format(datetime.datetime.now())),
description = "(automatically generated)"
)
time.sleep(10)
image_object = conn_ec2.get_image(new_image)
while image_object.state == "pending":
logging.info("State is still pending. Retrying in 10 seconds.")
time.sleep(10)
image_object.update()
finally:
logging.warn("Stopping all nodes.")
for node in reservation.instances:
node.terminate()
logging.info("Creating new LaunchConfiguration.")
mapping = BlockDeviceMapping()
mapping["/dev/sdb"] = BlockDeviceType(ephemeral_name = "ephemeral0")
mapping["/dev/sdc"] = BlockDeviceType(ephemeral_name = "ephemeral1")
new_launch_config = LaunchConfiguration(
conn_ec2_as,
name = ('Latest-{:%Y-%m-%d--%H-%M-%S}'.
format(datetime.datetime.now())),
image_id = new_image,
security_groups = ['sg-f9a08492'],
instance_type = 'c3.large',
block_device_mappings = [mapping],
instance_profile_name = ("arn:aws:iam::470084502640:instance-profile"
"/dal-access"),
spot_price = 0.02,
)
conn_ec2_as.create_launch_configuration(new_launch_config)
logging.info("Setting launch configuration in existing ASG.")
group.launch_config_name = new_launch_config.name
group.update()
logging.info("Cleaning up old launch configurations.")
for config in conn_ec2_as.get_all_launch_configurations():
if config.image_id != new_launch_config.image_id:
conn_ec2_as.delete_launch_configuration(config.name)
logging.info("Cleaning up old images.")
for image in conn_ec2.get_all_images(filters={"name":["LatestImage"]}):
if image.id != new_image:
conn_ec2.deregister_image(image.id, True)
logging.info("Rollout complete. New image is {}.".format(new_image))
if not options.no_restart:
logging.info("Triggering reload of all nodes in ASG.")
for instance in group.instances:
for reservation in conn_ec2.get_all_instances(instance.instance_id):
reservation.stop_all()
class WatchData:
datafile = "/var/tmp/watchdata.p"
dry = False
low_limit = 72
high_limit = 90
high_urgent = 95
stats_period = 60
history_size = 0
def __init__(self):
self.name = ''
self.instances = 0
self.new_desired = 0
self.desired = 0
self.instances_info = None
self.previous_instances = 0
self.action = ""
self.action_ts = 0
self.changed_ts = 0
self.total_load = 0
self.avg_load = 0
self.max_load = 0
self.up_ts = 0
self.down_ts = 0
self.max_loaded = None
self.loads = {}
self.measures = {}
self.emergency = False
self.history = None
self.trend = 0
self.exponential_average = 0
self.ts = 0
def __getstate__(self):
""" Don't store these objets """
d = self.__dict__.copy()
del d['ec2']
del d['cw']
del d['autoscale']
del d['group']
del d['instances_info']
return d
def connect(self, groupname):
self.ec2 = boto.connect_ec2()
self.cw = CloudWatchConnection()
self.autoscale = AutoScaleConnection()
self.group = self.autoscale.get_all_groups(names=[groupname])[0]
self.instances = len(self.group.instances)
self.desired = self.group.desired_capacity
self.name = groupname
self.ts = int(time.time())
def get_instances_info(self):
ids = [i.instance_id for i in self.group.instances]
self.instances_info = self.ec2.get_only_instances(instance_ids=ids)
def get_CPU_loads(self):
""" Read instances load and store in data """
measures = 0
for instance in self.group.instances:
load = self.get_instance_CPU_load(instance.instance_id)
if load is None:
continue
measures += 1
self.total_load += load
self.loads[instance.instance_id] = load
if load > self.max_load:
self.max_load = load
self.max_loaded = instance.instance_id
if measures > 0:
self.avg_load = self.total_load / measures
def get_instance_CPU_load(self, instance):
end = datetime.datetime.now()
start = end - datetime.timedelta(seconds=int(self.stats_period * 3))
m = self.cw.get_metric_statistics(self.stats_period, start, end,
"CPUUtilization", "AWS/EC2",
["Average"],
{"InstanceId": instance})
if len(m) > 0:
measures = self.measures[instance] = len(m)
ordered = sorted(m, key=lambda x: x['Timestamp'])
averages = [x['Average'] for x in ordered]
average = reduce(lambda x, y: 0.4 * x + 0.6 * y, averages[-2:])
return average
return None
@classmethod
def from_file(cls):
try:
data = pickle.load(open(cls.datafile, "rb"))
except:
data = WatchData()
return data
def store(self, annotation=False):
if self.history_size > 0:
if not self.history: self.history = []
self.history.append([
int(time.time()),
len(self.group.instances),
int(round(self.total_load)),
int(round(self.avg_load))
])
self.history = self.history[-self.history_size:]
pickle.dump(self, open(self.datafile, "wb"))
if annotation:
import utils
text = json.dumps(self.__getstate__(), skipkeys=True)
utils.store_annotation("ec2_watch", text)
def check_too_low(self):
for instance, load in self.loads.iteritems():
if load is not None and self.measures[
instance] > 1 and self.instances > 1 and load < self.avg_load * 0.2 and load < 4:
self.emergency = True
self.check_avg_low(
) # Check if the desired instanes can be decreased
self.action = "EMERGENCY LOW (%s %5.2f%%) " % (instance, load)
self.kill_instance(instance)
return True
return self.emergency
def check_too_high(self):
for instance, load in self.loads.iteritems():
if load is not None and self.measures[
instance] > 1 and load > self.high_urgent:
self.emergency = True
self.action = "EMERGENCY HIGH (%s %5.2f%%) " % (instance, load)
if self.instances > 1 and load > self.avg_load * 1.5:
self.action += " killing bad instance"
self.kill_instance(instance)
else:
self.action += " increasing instances to %d" % (
self.instances + 1, )
self.set_desired(self.instances + 1)
return True
return self.emergency
def check_avg_high(self):
threshold = self.high_limit
if self.instances == 1:
threshold = threshold * 0.9 # Increase faster if there is just one instance
if self.avg_load > threshold:
self.action = "WARN, high load: %d -> %d " % (self.instances,
self.instances + 1)
self.set_desired(self.instances + 1)
return True
def check_avg_low(self):
if self.instances <= self.group.min_size:
return False
if self.total_load / (self.instances - 1) < self.low_limit:
self.action = "low load: %d -> %d " % (self.instances,
self.instances - 1)
self.set_desired(self.instances - 1)
def kill_instance(self, id):
if self.action:
print(self.action)
print("Kill instance", id)
syslog.syslog(
syslog.LOG_INFO, "ec2_watch kill_instance: %s instances: %d (%s)" %
(id, self.instances, self.action))
if self.dry:
return
self.ec2.terminate_instances(instance_ids=[id])
self.action_ts = time.time()
def set_desired(self, desired):
if self.action:
print(self.action)
print("Setting instances from %d to %d" % (self.instances, desired))
syslog.syslog(
syslog.LOG_INFO, "ec2_watch set_desired: %d -> %d (%s)" %
(self.instances, desired, self.action))
if self.dry:
return
if desired >= self.group.min_size:
self.group.set_capacity(desired)
self.action_ts = time.time()
self.new_desired = desired
"--mothball",
type=str,
help="reduce the instances for all autoscaling group(s) "
"in a stack to zero")
parser.add_argument(
"-r",
"--reopen",
nargs=2,
help="increase the instances for all autoscaling group(s) "
"in a stack to min:max:desired")
args = parser.parse_args()
# connect to AWS
try:
cfn = CloudFormationConnection()
asg = AutoScaleConnection()
except:
print "AWS connect error"
else:
# get the key data
data = getStackAutoscalingGroupData(cfn, asg)
# list if explicitly listing or not doing anything else
if args.list or args.mothball is None and args.reopen is None:
for stackname in sorted(data, key=data.__getitem__):
print "{s}:".format(s=stackname, )
for asginfo in data[stackname]:
print " {n} {mn}:{mx}:{d}".format(
n=asginfo['name'],
mn=asginfo['min'],
mx=asginfo['max'],
d=asginfo['desired'],
class AutoScale:
def __init__(self, args):
"""
Initializing basic variables needed for auto scaling
"""
self.configs = ConfigParser.RawConfigParser()
self.args = args
self.test_props = {}
self.props = {}
self.ec2_connection = EC2Connection(self.args.access_key,
self.args.secret_key)
self.autoscale_connection = AutoScaleConnection(
self.args.access_key, self.args.secret_key)
self.elb_connection = ELBConnection(self.args.access_key,
self.args.secret_key)
self.cw_connection = CloudWatchConnection(self.args.access_key,
self.args.secret_key)
self.firstInstance = None
self.launchConfiguration = None
self.healthCheck = None
def loadConfigs(self):
"""
FIX ME: Currently doesnt do anything
This method will load the configurations from boto config file if present else will
accept parameters passed by user.
"""
if os.path.isfile("/etc/boto.cfg"):
self.configs.read("/etc/boto.cfg")
conf = self.configs.sections()
self.populateConfigs(conf)
if os.path.isfile("~/.boto"):
self.configs.read("~/.boto")
conf = self.configs.sections()
self.populateConfigs(conf)
print ">>> Loaded configs"
def populateConfigs(self, sections):
for section in sections:
self.boto_props[section] = self.configs.items(section)
for item in self.boto_props[section]:
key, value = item
if not self.props.has_key(key):
self.props[key] = value
def createLaunchConfiguration(self, lc_name, ami_id, key_name):
"""
Creates launch configuration for the auto scaling cluster
"""
self.launchConfiguration = LaunchConfiguration(name=lc_name,
image_id=ami_id,
key_name=key_name)
self.autoscale_connection.create_launch_configuration(
self.launchConfiguration)
print ">>> Created launch configuration: " + lc_name
def createAutoScaleGroup(self, asg_name):
"""
Create a Auto scaling group for the auto scaling cluster
"""
autoScalingGroup = AutoScalingGroup(
group_name=asg_name,
load_balancers=[self.args.lb_name],
launch_config=self.launchConfiguration,
min_size=self.args.min_size,
max_size=self.args.max_size,
availability_zones=['us-east-1a'])
self.autoscale_connection.create_auto_scaling_group(autoScalingGroup)
print ">>> Created auto scaling group: " + asg_name
def createTrigger(self, trigger_name, measure, asg_name):
"""
Trigger to spawn new instances as per specific metrics
"""
alarm_actions = []
dimensions = {"AutoScalingGroupName": asg_name}
policies = self.autoscale_connection.get_all_policies(
as_group=self.args.asg_name, policy_names=[self.args.asp_name])
for policy in policies:
alarm_actions.append(policy.policy_arn)
alarm = MetricAlarm(name=trigger_name,
namespace="AWS/EC2",
metric=measure,
statistic="Average",
comparison=">=",
threshold=50,
period=60,
unit="Percent",
evaluation_periods=2,
alarm_actions=alarm_actions,
dimensions=dimensions)
self.cw_connection.create_alarm(alarm)
print ">>> Created trigger: " + self.args.trigger
def createAutoScalePolicy(self, asp_name):
"""
Creates a Auto scaling policy to Add/Remove a instance from auto scaling cluster
"""
self.autoScalingUpPolicy = ScalingPolicy(
name=asp_name + '-up',
adjustment_type="ChangeInCapacity",
as_name=self.args.asg_name,
scaling_adjustment=1,
cooldown=180)
self.autoScalingDownPolicy = ScalingPolicy(
name=asp_name + '-down',
adjustment_type="ChangeInCapacity",
as_name=self.args.asg_name,
scaling_adjustment=-1,
cooldown=180)
self.autoscale_connection.create_scaling_policy(
self.autoScalingUpPolicy)
self.autoscale_connection.create_scaling_policy(
self.autoScalingDownPolicy)
print ">>> Created auto scaling policy: " + asp_name
def configureHealthCheck(self, target):
"""
Configures health check for the cluster
"""
self.healthCheck = HealthCheck(target=target, timeout=5)
print ">>> Configured health check for: " + target
def createLoadBalancer(self, lb_name, region, lb_port, instance_port,
protocol):
"""
Creates a load balancer for cluster
"""
listener = (int(lb_port), int(instance_port), protocol)
tuple_list = []
tuple_list.append(listener)
lbs = self.elb_connection.get_all_load_balancers()
for lb in lbs:
if lb.name != lb_name:
self.elb_connection.create_load_balancer(
lb_name, [region], tuple_list)
self.elb_connection.configure_health_check(
name=lb_name, health_check=self.healthCheck)
print ">>> Created load balancer: " + lb_name
else:
print "Load balancer with name '" + lb_name + "' already exists"
def startInstance(self, image_id, key_name, region, instance_type):
"""
Starts the first instance which will be serving requests irrespective of auto scaling
instances.
"""
reservation = self.ec2_connection.run_instances(
image_id=image_id,
min_count=1,
max_count=1,
placement=region,
key_name=key_name,
instance_type=instance_type)
# for instance in reservation.instances:
# instance.add_tag('node', '0')
# break
self.firstInstance = reservation.instances[0].id.split('\'')[0]
print ">>> Started instance: ", self.firstInstance
def registerInstanceToELB(self, lb_name):
"""
Register the first instance started to the Elastic Load Balancer.
"""
self.elb_connection.register_instances(load_balancer_name=lb_name,
instances=[self.firstInstance])
print ">>> Registered instance '", self.firstInstance, "' to load balancer '" + lb_name + "'"
def setUp(self):
"""
Set's up the auto scaling for the application
"""
# STEP 1: Load the configurations
self.loadConfigs()
# STEP 2: Configure the health check for the instances
self.configureHealthCheck(self.args.lb_target)
# STEP 3: Create a load balancer
self.createLoadBalancer(self.args.lb_name, self.args.region,
self.args.lb_port, self.args.instance_port,
self.args.protocol)
# STEP 4: Start the first instance
self.startInstance(self.args.ami_id, self.args.key_name,
self.args.region, self.args.instance_type)
# STEP 5: Register the instance to the load balancer created in STEP 4
self.registerInstanceToELB(self.args.lb_name)
# STEP 6: Create launch configuration to launch instances by auto scale
self.createLaunchConfiguration(self.args.lc_name, self.args.ami_id,
self.args.key_name)
# STEP 7: Create a auto scale group which will manage the instances started by auto scaling
self.createAutoScaleGroup(self.args.asg_name)
# STEP 8: Create a auto scaling policy to say add/remove a node
self.createAutoScalePolicy(self.args.asp_name)
# STEP 9: Create a trigger, so that auto scaling can trigger it to start
# or remove a instance from auto scaling group
self.createTrigger(self.args.trigger, self.args.measure,
self.args.asg_name)
class WatchData:
datafile = "/tmp/watchdata.p"
dry = False
low_limit = 70
high_limit = 90
high_urgent = 95
stats_period = 120
history_size = 0
def __init__(self):
self.name = ''
self.instances = 0
self.new_desired = 0
self.desired = 0
self.instances_info = None
self.previous_instances = 0
self.action = ""
self.action_ts = 0
self.changed_ts = 0
self.total_load = 0
self.avg_load = 0
self.max_load = 0
self.up_ts = 0
self.down_ts= 0
self.max_loaded = None
self.loads = {}
self.measures = {}
self.emergency = False
self.history = None
def __getstate__(self):
""" Don't store these objets """
d = self.__dict__.copy()
del d['ec2']
del d['cw']
del d['autoscale']
del d['group']
del d['instances_info']
return d
def connect(self, groupname):
self.ec2 = boto.connect_ec2()
self.cw = CloudWatchConnection()
self.autoscale = AutoScaleConnection()
self.group = self.autoscale.get_all_groups(names=[groupname])[0]
self.instances = len(self.group.instances)
self.desired = self.group.desired_capacity
self.name = groupname
def get_instances_info(self):
ids = [i.instance_id for i in self.group.instances]
self.instances_info = self.ec2.get_only_instances(instance_ids = ids)
def get_CPU_loads(self):
""" Read instances load and store in data """
for instance in self.group.instances:
load = self.get_instance_CPU_load(instance.instance_id)
if load is None:
continue
self.total_load += load
self.loads[instance.instance_id] = load
if load > self.max_load:
self.max_load = load
self.max_loaded = instance.instance_id
self.avg_load = self.total_load/self.instances
def get_instance_CPU_load(self, instance):
end = datetime.datetime.now()
start = end - datetime.timedelta(seconds=300)
m = self.cw.get_metric_statistics(self.stats_period, start, end, "CPUUtilization", "AWS/EC2", ["Average"], {"InstanceId": instance})
if len(m) > 0:
self.measures[instance] = len(m)
ordered = sorted(m, key=lambda x: x['Timestamp'], reverse=True)
return ordered[0]['Average']
return None
@classmethod
def from_file(cls):
try:
data = pickle.load( open(cls.datafile, "rb" ))
except:
data = WatchData()
return data
def store(self, annotation = False):
if self.history_size > 0:
if not self.history: self.history = []
self.history.append([int(time.time()), len(self.group.instances), int(round(self.total_load))])
self.history = self.history[-self.history_size:]
pickle.dump(self, open(self.datafile, "wb" ))
if annotation:
import utils
text = json.dumps(self.__getstate__(), skipkeys=True)
utils.store_annotation("ec2_watch", text)
def check_too_low(self):
for instance, load in self.loads.iteritems():
if load is not None and self.measures[instance] > 1 and self.instances > 1 and load < self.avg_load * 0.2 and load < 4:
self.emergency = True
self.check_avg_low() # Check if the desired instanes can be decreased
self.action = "EMERGENCY LOW (%s %5.2f%%) " % (instance, load)
self.kill_instance(instance)
return True
return self.emergency
def check_too_high(self):
for instance, load in self.loads.iteritems():
if load is not None and self.measures[instance] > 1 and load > self.high_urgent:
self.emergency = True
self.action = "EMERGENCY HIGH (%s %5.2f%%) " % (instance, load)
if self.instances > 1 and load > self.avg_load * 1.5:
self.action += " killing bad instance"
self.kill_instance(instance)
else:
self.action += " increasing instances to %d" % (self.instances+1,)
self.set_desired(self.instances+1)
return True
return self.emergency
def check_avg_high(self):
threshold = self.high_limit
if self.instances == 1:
threshold = threshold * 0.9 # Increase faster if there is just one instance
if self.avg_load > threshold:
self.action = "WARN, high load: %d -> %d " % (self.instances, self.instances + 1)
self.set_desired(self.instances + 1)
return True
def check_avg_low(self):
if self.instances <= self.group.min_size:
return False
if self.total_load/(self.instances-1) < self.low_limit:
self.action = "low load: %d -> %d " % (self.instances, self.instances - 1)
self.set_desired(self.instances - 1)
def kill_instance(self, id):
if self.action:
print self.action
print "Kill instance", id
syslog.syslog(syslog.LOG_INFO, "ec2_watch kill_instance: %s instances: %d (%s)" % (id, self.instances, self.action))
if self.dry:
return
self.ec2.terminate_instances(instance_ids=[id])
self.action_ts = time.time()
def set_desired(self, desired):
if self.action:
print self.action
print "Setting instances from %d to %d" % (self.instances, desired)
syslog.syslog(syslog.LOG_INFO, "ec2_watch set_desired: %d -> %d (%s)" % (self.instances, desired, self.action))
if self.dry:
return
if desired >= self.group.min_size:
self.group.set_capacity(desired)
self.action_ts = time.time()
self.new_desired = desired
def create_autoscaling(ami_id, sns_arn):
"""
Creates the autoscaling group for proxy instances
Inspired by boto autoscaling tutorial.
"""
con = AutoScaleConnection(aws_secret_access_key=AWS_SECRET_ACCESS_KEY,
aws_access_key_id=AWS_ACCESS_KEY,
region=RegionInfo(name=REGION,
endpoint='autoscaling.%s.amazonaws.com' % REGION))
print "Creating autoscaling configuration.."
config = LaunchConfiguration(name=AUTOSCALING_GROUP_NAME,
image_id=ami_id,
key_name=KEY_NAME,
security_groups=[EC2_SECURITY_GROUP_NAME],
instance_type=INSTANCE_TYPE)
con.create_launch_configuration(config)
print "Create autoscaling group..."
ag = AutoScalingGroup(name=AUTOSCALING_GROUP_NAME,
launch_config=config,
availability_zones=["{0}a".format(REGION)],
load_balancers=[ELB_NAME],
min_size=AUTOSCALING_MIN_INSTANCES,
max_size=AUTOSCALING_MAX_INSTANCES,
group_name=AUTOSCALING_GROUP_NAME)
con.create_auto_scaling_group(ag)
# fetch the autoscale group after it is created (unused but may be necessary)
_ = con.get_all_groups(names=[AUTOSCALING_GROUP_NAME])[0]
# Create tag name for autoscaling-created machines
as_tag = Tag(key='Name', value=AUTOSCALING_GROUP_NAME, propagate_at_launch=True, resource_id=AUTOSCALING_GROUP_NAME)
con.create_or_update_tags([as_tag])
print "Creating autoscaling policy..."
scaleup_policy = ScalingPolicy(name='scale_up',
adjustment_type='ChangeInCapacity',
as_name=AUTOSCALING_GROUP_NAME,
scaling_adjustment=1,
cooldown=AUTOSCALING_COOLDOWN_PERIOD)
scaledown_policy = ScalingPolicy(name='scale_down',
adjustment_type='ChangeInCapacity',
as_name=AUTOSCALING_GROUP_NAME,
scaling_adjustment=-1,
cooldown=AUTOSCALING_COOLDOWN_PERIOD)
con.create_scaling_policy(scaleup_policy)
con.create_scaling_policy(scaledown_policy)
# Get freshened policy objects
scaleup_policy = con.get_all_policies(as_group=AUTOSCALING_GROUP_NAME, policy_names=['scale_up'])[0]
scaledown_policy = con.get_all_policies(as_group=AUTOSCALING_GROUP_NAME, policy_names=['scale_down'])[0]
print "Creating cloudwatch alarms"
cloudwatch_con = CloudWatchConnection(aws_secret_access_key=AWS_SECRET_ACCESS_KEY,
aws_access_key_id=AWS_ACCESS_KEY,
region=RegionInfo(name=REGION,
endpoint='monitoring.%s.amazonaws.com' % REGION))
alarm_dimensions = {"AutoScalingGroupName": AUTOSCALING_GROUP_NAME}
scaleup_alarm = MetricAlarm(name='scale_up_on_cpu',
namespace='AWS/EC2',
metric='CPUUtilization',
statistic='Average',
comparison='>',
threshold=AUTOSCALING_CPU_MAX_THRESHOLD,
period='60',
evaluation_periods=1,
alarm_actions=[scaleup_policy.policy_arn, sns_arn],
dimensions=alarm_dimensions)
# Don't send SNS on scaledown policy
scaledown_alarm = MetricAlarm(name='scale_down_on_cpu',
namespace='AWS/EC2',
metric='CPUUtilization',
statistic='Average',
comparison='<',
threshold=AUTOSCALING_CPU_MIN_THRESHOLD,
period='60',
evaluation_periods=1,
alarm_actions=[scaledown_policy.policy_arn],
dimensions=alarm_dimensions)
cloudwatch_con.create_alarm(scaleup_alarm)
cloudwatch_con.create_alarm(scaledown_alarm)
def test_basic(self):
# NB: as it says on the tin these are really basic tests that only
# (lightly) exercise read-only behaviour - and that's only if you
# have any autoscale groups to introspect. It's useful, however, to
# catch simple errors
print('--- running %s tests ---' % self.__class__.__name__)
c = AutoScaleConnection()
self.assertTrue(repr(c).startswith('AutoScaleConnection'))
groups = c.get_all_groups()
for group in groups:
self.assertIsInstance(group, AutoScalingGroup)
# get activities
activities = group.get_activities()
for activity in activities:
self.assertIsInstance(activity, Activity)
# get launch configs
configs = c.get_all_launch_configurations()
for config in configs:
self.assertIsInstance(config, LaunchConfiguration)
# get policies
policies = c.get_all_policies()
for policy in policies:
self.assertIsInstance(policy, ScalingPolicy)
# get scheduled actions
actions = c.get_all_scheduled_actions()
for action in actions:
self.assertIsInstance(action, ScheduledUpdateGroupAction)
# get instances
instances = c.get_all_autoscaling_instances()
for instance in instances:
self.assertIsInstance(instance, Instance)
# get all scaling process types
ptypes = c.get_all_scaling_process_types()
for ptype in ptypes:
self.assertTrue(ptype, ProcessType)
# get adjustment types
adjustments = c.get_all_adjustment_types()
for adjustment in adjustments:
self.assertIsInstance(adjustment, AdjustmentType)
# get metrics collection types
types = c.get_all_metric_collection_types()
self.assertIsInstance(types, MetricCollectionTypes)
# create the simplest possible AutoScale group
# first create the launch configuration
time_string = '%d' % int(time.time())
lc_name = 'lc-%s' % time_string
lc = LaunchConfiguration(name=lc_name, image_id='ami-2272864b',
instance_type='t1.micro')
c.create_launch_configuration(lc)
found = False
lcs = c.get_all_launch_configurations()
for lc in lcs:
if lc.name == lc_name:
found = True
break
assert found
# now create autoscaling group
group_name = 'group-%s' % time_string
group = AutoScalingGroup(name=group_name, launch_config=lc,
availability_zones=['us-east-1a'],
min_size=1, max_size=1)
c.create_auto_scaling_group(group)
found = False
groups = c.get_all_groups()
for group in groups:
if group.name == group_name:
found = True
break
assert found
# now create a tag
tag = Tag(key='foo', value='bar', resource_id=group_name,
propagate_at_launch=True)
c.create_or_update_tags([tag])
found = False
tags = c.get_all_tags()
for tag in tags:
if tag.resource_id == group_name and tag.key == 'foo':
found = True
break
assert found
c.delete_tags([tag])
# shutdown instances and wait for them to disappear
group.shutdown_instances()
instances = True
while instances:
time.sleep(5)
groups = c.get_all_groups()
for group in groups:
if group.name == group_name:
if not group.instances:
instances = False
group.delete()
lc.delete()
found = True
while found:
found = False
time.sleep(5)
tags = c.get_all_tags()
for tag in tags:
if tag.resource_id == group_name and tag.key == 'foo':
found = True
assert not found
print('--- tests completed ---')
class BotoScaleInterface(ScaleInterface):
conn = None
saveclcdata = False
def __init__(self, clc_host, access_id, secret_key, token):
#boto.set_stream_logger('foo')
path='/services/AutoScaling'
port=8773
if clc_host[len(clc_host)-13:] == 'amazonaws.com':
clc_host = clc_host.replace('ec2', 'autoscaling', 1)
path = '/'
reg = None
port=443
reg = RegionInfo(name='eucalyptus', endpoint=clc_host)
self.conn = AutoScaleConnection(access_id, secret_key, region=reg,
port=port, path=path,
is_secure=True, security_token=token, debug=0)
self.conn.https_validate_certificates = False
self.conn.http_connection_kwargs['timeout'] = 30
def __save_json__(self, obj, name):
f = open(name, 'w')
json.dump(obj, f, cls=BotoJsonScaleEncoder, indent=2)
f.close()
##
# autoscaling methods
##
def create_auto_scaling_group(self, as_group):
return self.conn.create_auto_scaling_group(as_group)
def delete_auto_scaling_group(self, name, force_delete=False):
return self.conn.delete_auto_scaling_group(name, force_delete)
def get_all_groups(self, names=None, max_records=None, next_token=None):
obj = self.conn.get_all_groups(names, max_records, next_token)
if self.saveclcdata:
self.__save_json__(obj, "mockdata/AS_Groups.json")
return obj
def get_all_autoscaling_instances(self, instance_ids=None, max_records=None, next_token=None):
obj = self.conn.get_all_autoscaling_instances(instance_ids, max_records, next_token)
if self.saveclcdata:
self.__save_json__(obj, "mockdata/AS_Instances.json")
return obj
def set_desired_capacity(self, group_name, desired_capacity, honor_cooldown=False):
group = self.conn.get_all_groups([group_name])[0];
# notice, honor_cooldown not supported.
return group.set_capacity(desired_capacity)
def set_instance_health(self, instance_id, health_status, should_respect_grace_period=True):
return self.conn.set_instance_health(instance_id, health_status,
should_respect_grace_period)
def terminate_instance(self, instance_id, decrement_capacity=True):
return self.conn.terminate_instance(instance_id, decrement_capacity)
def update_autoscaling_group(self, as_group):
as_group.connection = self.conn
return as_group.update()
def create_launch_configuration(self, launch_config):
return self.conn.create_launch_configuration(launch_config)
def delete_launch_configuration(self, launch_config_name):
return self.conn.delete_launch_configuration(launch_config_name)
def get_all_launch_configurations(self, config_names, max_records, next_token):
obj = self.conn.get_all_launch_configurations(names=config_names, max_records=max_records, next_token=next_token)
if self.saveclcdata:
self.__save_json__(obj, "mockdata/AS_LaunchConfigs.json")
return obj
class BotoScaleInterface(ScaleInterface):
conn = None
saveclcdata = False
def __init__(self, clc_host, access_id, secret_key, token):
self.access_id = access_id
self.secret_key = secret_key
self.token = token
self.set_endpoint(clc_host)
def set_endpoint(self, endpoint):
#boto.set_stream_logger('scale')
path = '/services/AutoScaling'
reg = RegionInfo(name='eucalyptus', endpoint=endpoint)
port = 8773
if endpoint[len(endpoint)-13:] == 'amazonaws.com':
endpoint = endpoint.replace('ec2', 'autoscaling', 1)
path = '/'
reg = RegionInfo(endpoint=endpoint)
port = 443
self.conn = AutoScaleConnection(self.access_id, self.secret_key, region=reg,
port=port, path=path,
is_secure=True, security_token=self.token, debug=0)
self.conn.APIVersion = '2011-01-01'
if not(endpoint[len(endpoint)-13:] == 'amazonaws.com'):
self.conn.auth_region_name = 'Eucalyptus'
self.conn.https_validate_certificates = False
self.conn.http_connection_kwargs['timeout'] = 30
def __save_json__(self, obj, name):
f = open(name, 'w')
json.dump(obj, f, cls=BotoJsonScaleEncoder, indent=2)
f.close()
##
# autoscaling methods
##
def create_auto_scaling_group(self, as_group):
return self.conn.create_auto_scaling_group(as_group)
def delete_auto_scaling_group(self, name, force_delete=False):
return self.conn.delete_auto_scaling_group(name, force_delete)
def get_all_groups(self, names=None, max_records=None, next_token=None):
obj = self.conn.get_all_groups(names, max_records, next_token)
if self.saveclcdata:
self.__save_json__(obj, "mockdata/AS_Groups.json")
return obj
def get_all_autoscaling_instances(self, instance_ids=None, max_records=None, next_token=None):
obj = self.conn.get_all_autoscaling_instances(instance_ids, max_records, next_token)
if self.saveclcdata:
self.__save_json__(obj, "mockdata/AS_Instances.json")
return obj
def set_desired_capacity(self, group_name, desired_capacity, honor_cooldown=False):
group = self.conn.get_all_groups([group_name])[0];
# notice, honor_cooldown not supported.
return group.set_capacity(desired_capacity)
def set_instance_health(self, instance_id, health_status, should_respect_grace_period=True):
return self.conn.set_instance_health(instance_id, health_status,
should_respect_grace_period)
def terminate_instance(self, instance_id, decrement_capacity=True):
return self.conn.terminate_instance(instance_id, decrement_capacity)
def update_autoscaling_group(self, as_group):
as_group.connection = self.conn
return as_group.update()
def create_launch_configuration(self, launch_config):
return self.conn.create_launch_configuration(launch_config)
def delete_launch_configuration(self, launch_config_name):
return self.conn.delete_launch_configuration(launch_config_name)
def get_all_launch_configurations(self, config_names=None, max_records=None, next_token=None):
obj = self.conn.get_all_launch_configurations(names=config_names, max_records=max_records,
next_token=next_token)
if self.saveclcdata:
self.__save_json__(obj, "mockdata/AS_LaunchConfigs.json")
return obj
# policy related
def delete_policy(self, policy_name, autoscale_group=None):
return self.conn.delete_policy(policy_name, autoscale_group)
def get_all_policies(self, as_group=None, policy_names=None, max_records=None, next_token=None):
obj = self.conn.get_all_policies(as_group, policy_names, max_records, next_token)
if self.saveclcdata:
self.__save_json__(obj, "mockdata/AS_Policies.json")
return obj
def execute_policy(self, policy_name, as_group=None, honor_cooldown=None):
return self.conn.execute_policy(policy_name, as_group, honor_cooldown)
def create_scaling_policy(self, scaling_policy):
return self.conn.create_scaling_policy(scaling_policy)
def get_all_adjustment_types(self):
return self.conn.get_all_adjustment_types()
# tag related
def delete_tags(self, tags):
return self.conn.delete_tags(tags)
def get_all_tags(self, filters=None, max_records=None, next_token=None):
obj = self.conn.get_all_tags(filters, max_records, next_token)
if self.saveclcdata:
self.__save_json__(obj, "mockdata/AS_Tags.json")
return obj
def create_or_update_tags(self, tags):
return self.conn.create_or_update_tags(tags)
class MSBManager:
def __init__(self, aws_access_key, aws_secret_key):
self.ec2_conn = EC2Connection(aws_access_key, aws_secret_key)
self.elb_conn = ELBConnection(aws_access_key, aws_secret_key)
self.auto_scale_conn = AutoScaleConnection(aws_access_key, aws_secret_key)
self.cloud_watch_conn = CloudWatchConnection(aws_access_key, aws_secret_key)
self.default_cooldown = 60
def get_security_group(self, name):
sgs = [g for g in self.ec2_conn.get_all_security_groups() if g.name == name]
return sgs[0] if sgs else None
def create_security_group(self, name, description):
sgs = [g for g in self.ec2_conn.get_all_security_groups() if g.name == name]
sg = sgs[0] if sgs else None
if not sgs:
sg = self.ec2_conn.create_security_group(name, description)
try:
sg.authorize(ip_protocol="-1", from_port=None, to_port=None, cidr_ip="0.0.0.0/0", dry_run=False)
except EC2ResponseError:
pass
return sg
def remove_security_group(self, name):
self.ec2_conn.delete_security_group(name=name)
def create_instance(self, image, instance_type, key_name, zone, security_groups, tags):
instance = None
reservations = self.ec2_conn.get_all_instances()
for reservation in reservations:
for i in reservation.instances:
if "Name" in i.tags and i.tags["Name"] == tags["Name"] and i.state == "running":
instance = i
break
if not instance:
reservation = self.ec2_conn.run_instances(
image,
instance_type=instance_type,
key_name=key_name,
placement=zone,
security_groups=security_groups,
monitoring_enabled=True,
)
instance = reservation.instances[0]
while not instance.update() == "running":
time.sleep(5)
time.sleep(10)
self.ec2_conn.create_tags([instance.id], tags)
return instance
def request_spot_instance(self, bid, image, instance_type, key_name, zone, security_groups, tags):
req = self.ec2_conn.request_spot_instances(
price=bid,
instance_type=instance_type,
image_id=image,
placement=zone,
key_name=key_name,
security_groups=security_groups,
)
instance_id = None
while not instance_id:
job_sir_id = req[0].id
requests = self.ec2_conn.get_all_spot_instance_requests()
for sir in requests:
if sir.id == job_sir_id:
instance_id = sir.instance_id
break
print "Job {} not ready".format(job_sir_id)
time.sleep(60)
self.ec2_conn.create_tags([instance_id], tags)
def remove_instance(self, instance_id):
self.remove_instances([instance_id])
def remove_instances(self, instance_ids):
self.ec2_conn.terminate_instances(instance_ids)
def remove_instance_by_tag_name(self, name):
reservations = self.ec2_conn.get_all_instances()
data_centers_intance_ids = []
for reservation in reservations:
for instance in reservation.instances:
if "Name" in instance.tags and instance.tags["Name"] == name and instance.state == "running":
data_centers_intance_ids.append(instance.id)
if data_centers_intance_ids:
self.remove_instances(data_centers_intance_ids)
def create_elb(self, name, zone, project_tag_value, security_group_id, instance_ids=None):
lbs = [l for l in self.elb_conn.get_all_load_balancers() if l.name == name]
lb = lbs[0] if lbs else None
if not lb:
hc = HealthCheck(
timeout=50, interval=60, healthy_threshold=2, unhealthy_threshold=8, target="HTTP:80/heartbeat"
)
ports = [(80, 80, "http")]
zones = [zone]
lb = self.elb_conn.create_load_balancer(name, zones, ports)
self.elb_conn.apply_security_groups_to_lb(name, [security_group_id])
lb.configure_health_check(hc)
if instance_ids:
lb.register_instances(instance_ids)
params = {
"LoadBalancerNames.member.1": lb.name,
"Tags.member.1.Key": "15619project",
"Tags.member.1.Value": project_tag_value,
}
lb.connection.get_status("AddTags", params, verb="POST")
return lb
def remove_elb(self, name):
self.elb_conn.delete_load_balancer(name)
def create_launch_configuration(self, name, image, key_name, security_groups, instance_type):
lcs = [l for l in self.auto_scale_conn.get_all_launch_configurations() if l.name == name]
lc = lcs[0] if lcs else None
if not lc:
lc = LaunchConfiguration(
name=name,
image_id=image,
key_name=key_name,
security_groups=[security_groups],
instance_type=instance_type,
)
self.auto_scale_conn.create_launch_configuration(lc)
return lc
def remove_launch_configuration(self, name):
self.auto_scale_conn.delete_launch_configuration(name)
def create_autoscaling_group(self, name, lb_name, zone, tags, instance_ids=None):
lc = self.create_launch_configuration()
as_groups = [a for a in self.auto_scale_conn.get_all_groups() if a.name == name]
as_group = as_groups[0] if as_groups else None
if not as_group:
as_group = AutoScalingGroup(
group_name=name,
load_balancers=[lb_name],
availability_zones=[zone],
launch_config=lc,
min_size=4,
max_size=4,
health_check_type="ELB",
health_check_period=120,
connection=self.auto_scale_conn,
default_cooldown=self.default_cooldown,
desired_capacity=4,
tags=tags,
)
self.auto_scale_conn.create_auto_scaling_group(as_group)
if instance_ids:
self.auto_scale_conn.attach_instances(name, instance_ids)
scale_up_policy = ScalingPolicy(
name="scale_up",
adjustment_type="ChangeInCapacity",
as_name=name,
scaling_adjustment=1,
cooldown=self.default_cooldown,
)
scale_down_policy = ScalingPolicy(
name="scale_down",
adjustment_type="ChangeInCapacity",
as_name=name,
scaling_adjustment=-1,
cooldown=self.default_cooldown,
)
self.auto_scale_conn.create_scaling_policy(scale_up_policy)
self.auto_scale_conn.create_scaling_policy(scale_down_policy)
scale_up_policy = self.auto_scale_conn.get_all_policies(as_group=name, policy_names=["scale_up"])[0]
scale_down_policy = self.auto_scale_conn.get_all_policies(as_group=name, policy_names=["scale_down"])[0]
alarm_dimensions = {"AutoScalingGroupName": name}
scale_up_alarm = MetricAlarm(
name="scale_up_on_cpu",
namespace="AWS/EC2",
metric="CPUUtilization",
statistic="Average",
comparison=">",
threshold=85,
period=60,
evaluation_periods=1,
alarm_actions=[scale_up_policy.policy_arn],
dimensions=alarm_dimensions,
)
self.cloud_watch_conn.create_alarm(scale_up_alarm)
scale_down_alarm = MetricAlarm(
name="scale_down_on_cpu",
namespace="AWS/EC2",
metric="CPUUtilization",
statistic="Average",
comparison="<",
threshold=60,
period=60,
evaluation_periods=1,
alarm_actions=[scale_down_policy.policy_arn],
dimensions=alarm_dimensions,
)
self.cloud_watch_conn.create_alarm(scale_down_alarm)
return as_group
def update_autoscaling_group_max_size(self, as_group, max_size):
setattr(as_group, "max_size", max_size)
as_group.update()
def update_autoscaling_group_min_size(self, as_group, min_size):
setattr(as_group, "min_size", min_size)
as_group.update()
def remove_autoscaling_group(self, name):
self.auto_scale_conn.delete_auto_scaling_group(name)
def add_ingress_rule(dry_run, go_agent_security_group, go_agent_security_group_owner, go_agent_security_group_name):
"""
For each ASG (app) in each VPC, add a rule to each SG associated with the ASG's launch configuration
that allows SSH ingress from the GoCD agents' SG.
BEFORE RUNNING THIS SCRIPT!:
- Use the assume_role bash script to assume the role in the proper account/VPC (edx, edge, mckinsey, etc.)
- If you don't know what this is, ask someone in DevOps.
- THEN run this script.
"""
asg_conn = AutoScaleConnection()
ec2_conn = boto.ec2.connect_to_region('us-east-1')
asgs = []
launch_configs = {}
security_groups = {}
logging.debug('All ASGs:')
for group in asg_conn.get_all_groups():
logging.debug(' {}'.format(group))
asgs.append(group)
logging.debug('All launch configurations:')
for launch_config in asg_conn.get_all_launch_configurations():
logging.debug(' {}'.format(launch_config))
launch_configs[launch_config.name] = launch_config
logging.debug('All security groups:')
for sec_group in ec2_conn.get_all_security_groups():
logging.debug(' {}'.format(sec_group))
security_groups[sec_group.id] = sec_group
# Validate that each ASG has a launch configuration.
for group in asgs:
try:
logging.info("Launch configuration for ASG '{}' is '{}'.".format(
group.name, launch_configs[group.launch_config_name]
))
except KeyError:
logging.error("Launch configuration '{}' for ASG '{}' was not found!".format(
group.launch_config_name, group.name
))
raise
# Construct a fake security group for the prod-tools-goagent-sg security group in the edx-tools account.
# This group will be used to grant the go-agents ingress into the ASG's VPCs.
go_agent_security_group = boto.ec2.securitygroup.SecurityGroup(
name=go_agent_security_group_name,
owner_id=go_agent_security_group_owner,
id=go_agent_security_group
)
# For each launch config, check for the security group. Can support multiple security groups
# but the edX DevOps convention is to use a single security group.
for group in asgs:
launch_config = launch_configs[group.launch_config_name]
if len(launch_config.security_groups) > 1:
err_msg = "Launch config '{}' for ASG '{}' has more than one security group!: {}".format(
launch_config.name, group.name, launch_config.security_groups
)
logging.warning(err_msg)
continue
sg_name = launch_config.security_groups[0]
try:
# Find the security group.
sec_group = security_groups[sg_name]
except KeyError:
logging.error("Security group '{}' for ASG '{}' was not found!.".format(sg_name, group.name))
logging.info('BEFORE: Rules for security group {}:'.format(sec_group.name))
logging.info(sec_group.rules)
try:
# Add the ingress rule to the security group.
yes_no = six.moves.input("Apply the change to this security group? [Yes]")
if yes_no in ("", "y", "Y", "yes"):
sec_group.authorize(
ip_protocol='tcp',
from_port=22,
to_port=22,
src_group=go_agent_security_group,
dry_run=dry_run
)
except boto.exception.EC2ResponseError as exc:
if exc.status == 412:
# If the dry_run flag is set, then each rule addition will raise this exception.
# Log it and carry on.
logging.info('Dry run is True but rule addition would have succeeded for security group {}.'.format(
sg_name
))
elif exc.code == "InvalidPermission.Duplicate":
logging.info("Rule already exists for {}.".format(sg_name))
else:
raise
logging.info('AFTER: Rules for security group {}:'.format(sg_name))
logging.info(sec_group.rules)
def add_ingress_rule(dry_run, go_agent_security_group, go_agent_security_group_owner, go_agent_security_group_name):
"""
For each ASG (app) in each VPC, add a rule to each SG associated with the ASG's launch configuration
that allows SSH ingress from the GoCD agents' SG.
BEFORE RUNNING THIS SCRIPT!:
- Use the assume_role bash script to assume the role in the proper account/VPC (edx, edge, mckinsey, etc.)
- If you don't know what this is, ask someone in DevOps.
- THEN run this script.
"""
asg_conn = AutoScaleConnection()
ec2_conn = boto.ec2.connect_to_region('us-east-1')
asgs = []
launch_configs = {}
security_groups = {}
logging.debug('All ASGs:')
for group in asg_conn.get_all_groups():
logging.debug(' {}'.format(group))
asgs.append(group)
logging.debug('All launch configurations:')
for launch_config in asg_conn.get_all_launch_configurations():
logging.debug(' {}'.format(launch_config))
launch_configs[launch_config.name] = launch_config
logging.debug('All security groups:')
for sec_group in ec2_conn.get_all_security_groups():
logging.debug(' {}'.format(sec_group))
security_groups[sec_group.id] = sec_group
# Validate that each ASG has a launch configuration.
for group in asgs:
try:
logging.info("Launch configuration for ASG '{}' is '{}'.".format(
group.name, launch_configs[group.launch_config_name]
))
except KeyError:
logging.error("Launch configuration '{}' for ASG '{}' was not found!".format(
group.launch_config_name, group.name
))
raise
# Construct a fake security group for the prod-tools-goagent-sg security group in the edx-tools account.
# This group will be used to grant the go-agents ingress into the ASG's VPCs.
go_agent_security_group = boto.ec2.securitygroup.SecurityGroup(
name=go_agent_security_group_name,
owner_id=go_agent_security_group_owner,
id=go_agent_security_group
)
# For each launch config, check for the security group. Can support multiple security groups
# but the edX DevOps convention is to use a single security group.
for group in asgs:
launch_config = launch_configs[group.launch_config_name]
if len(launch_config.security_groups) > 1:
err_msg = "Launch config '{}' for ASG '{}' has more than one security group!: {}".format(
launch_config.name, group.name, launch_config.security_groups
)
logging.warning(err_msg)
continue
sg_name = launch_config.security_groups[0]
try:
# Find the security group.
sec_group = security_groups[sg_name]
except KeyError:
logging.error("Security group '{}' for ASG '{}' was not found!.".format(sg_name, group.name))
logging.info('BEFORE: Rules for security group {}:'.format(sec_group.name))
logging.info(sec_group.rules)
try:
# Add the ingress rule to the security group.
yes_no = raw_input("Apply the change to this security group? [Yes]")
if yes_no in ("", "y", "Y", "yes"):
sec_group.authorize(
ip_protocol='tcp',
from_port=22,
to_port=22,
src_group=go_agent_security_group,
dry_run=dry_run
)
except boto.exception.EC2ResponseError as exc:
if exc.status == 412:
# If the dry_run flag is set, then each rule addition will raise this exception.
# Log it and carry on.
logging.info('Dry run is True but rule addition would have succeeded for security group {}.'.format(
sg_name
))
elif exc.code == "InvalidPermission.Duplicate":
logging.info("Rule already exists for {}.".format(sg_name))
else:
raise
logging.info('AFTER: Rules for security group {}:'.format(sg_name))
logging.info(sec_group.rules)
healthy_threshold=10,
unhealthy_threshold=2,
target=elb['target'],
timeout=elb['time_out'])
lb.configure_health_check(hc)
params = {"LoadBalancerNames.member.1": lb.name,
"Tags.member.1.Key": 'Project',
"Tags.member.1.Value": '2.2'}
s=lb.connection.get_status('AddTags', params, verb='POST')
print 'Load Balancer DNS: ' + lb_dns
# -------------------------Create Auto Scaling Group----------------------------
con_as = AutoScaleConnection()
lc = LaunchConfiguration(name='Project2.2_Lauch_Config',
image_id=DC_IMAGE,
key_name=KEY_NAME,
security_groups=SECURITY_GROUP2,
instance_type=DC_INSTANCE_TYPE,
instance_monitoring=DETAIL_MON)
con_as.create_launch_configuration(lc)
asg = AutoScalingGroup(name='Project2.2_AutoSacling_Group',
load_balancers=[elb['name']],
availability_zones=ZONE,
health_check_period='120',
health_check_type='ELB',
launch_config=lc,
min_size=1,
def __init__(self, aws_access_key, aws_secret_key):
self.ec2_conn = EC2Connection(aws_access_key, aws_secret_key)
self.elb_conn = ELBConnection(aws_access_key, aws_secret_key)
self.auto_scale_conn = AutoScaleConnection(aws_access_key, aws_secret_key)
self.cloud_watch_conn = CloudWatchConnection(aws_access_key, aws_secret_key)
self.default_cooldown = 60
def __init__(self, region, **kwargs):
self.conn = AutoScaleConnection(region=get_region(region), **kwargs)
class BotoScaleInterface(ScaleInterface):
conn = None
saveclcdata = False
def __init__(self, clc_host, access_id, secret_key, token):
#boto.set_stream_logger('foo')
path = '/services/AutoScaling'
reg = RegionInfo(name='eucalyptus', endpoint=clc_host)
port = 8773
if clc_host[len(clc_host) - 13:] == 'amazonaws.com':
clc_host = clc_host.replace('ec2', 'autoscaling', 1)
path = '/'
reg = None
port = 443
self.conn = AutoScaleConnection(access_id,
secret_key,
region=reg,
port=port,
path=path,
is_secure=True,
security_token=token,
debug=0)
self.conn.APIVersion = '2011-01-01'
if not (clc_host[len(clc_host) - 13:] == 'amazonaws.com'):
self.conn.auth_region_name = 'Eucalyptus'
self.conn.https_validate_certificates = False
self.conn.http_connection_kwargs['timeout'] = 30
def __save_json__(self, obj, name):
f = open(name, 'w')
json.dump(obj, f, cls=BotoJsonScaleEncoder, indent=2)
f.close()
##
# autoscaling methods
##
def create_auto_scaling_group(self, as_group):
return self.conn.create_auto_scaling_group(as_group)
def delete_auto_scaling_group(self, name, force_delete=False):
return self.conn.delete_auto_scaling_group(name, force_delete)
def get_all_groups(self, names=None, max_records=None, next_token=None):
obj = self.conn.get_all_groups(names, max_records, next_token)
if self.saveclcdata:
self.__save_json__(obj, "mockdata/AS_Groups.json")
return obj
def get_all_autoscaling_instances(self,
instance_ids=None,
max_records=None,
next_token=None):
obj = self.conn.get_all_autoscaling_instances(instance_ids,
max_records, next_token)
if self.saveclcdata:
self.__save_json__(obj, "mockdata/AS_Instances.json")
return obj
def set_desired_capacity(self,
group_name,
desired_capacity,
honor_cooldown=False):
group = self.conn.get_all_groups([group_name])[0]
# notice, honor_cooldown not supported.
return group.set_capacity(desired_capacity)
def set_instance_health(self,
instance_id,
health_status,
should_respect_grace_period=True):
return self.conn.set_instance_health(instance_id, health_status,
should_respect_grace_period)
def terminate_instance(self, instance_id, decrement_capacity=True):
return self.conn.terminate_instance(instance_id, decrement_capacity)
def update_autoscaling_group(self, as_group):
as_group.connection = self.conn
return as_group.update()
def create_launch_configuration(self, launch_config):
return self.conn.create_launch_configuration(launch_config)
def delete_launch_configuration(self, launch_config_name):
return self.conn.delete_launch_configuration(launch_config_name)
def get_all_launch_configurations(self,
config_names=None,
max_records=None,
next_token=None):
obj = self.conn.get_all_launch_configurations(names=config_names,
max_records=max_records,
next_token=next_token)
if self.saveclcdata:
self.__save_json__(obj, "mockdata/AS_LaunchConfigs.json")
return obj
# policy related
def delete_policy(self, policy_name, autoscale_group=None):
return self.conn.delete_policy(policy_name, autoscale_group)
def get_all_policies(self,
as_group=None,
policy_names=None,
max_records=None,
next_token=None):
return self.conn.get_all_policies(as_group, policy_names, max_records,
next_token)
def execute_policy(self, policy_name, as_group=None, honor_cooldown=None):
return self.conn.execute_policy(policy_name, as_group, honor_cooldown)
def create_scaling_policy(self, scaling_policy):
return self.conn.create_scaling_policy(scaling_policy)
def get_all_adjustment_types(self):
return self.conn.get_all_adjustment_types()
# tag related
def delete_tags(self, tags):
return self.conn.delete_tags(tags)
def get_all_tags(self, filters=None, max_records=None, next_token=None):
return self.conn.get_all_tags(filters, max_records, next_token)
def create_or_update_tags(self, tags):
return self.conn.create_or_update_tags(tags)
'instance_monitoring':
True #Indicated whether the instances will be launched with detailed monitoring enabled. Needed to enable CloudWatch
}
##############################END CONFIGURATION#######################################
#=================Construct a list of all availability zones for your region=========
conn_reg = boto.ec2.connect_to_region(region_name=region)
zones = conn_reg.get_all_zones()
zoneStrings = []
for zone in zones:
zoneStrings.append(zone.name)
conn_elb = ELBConnection(AWS_ACCESS_KEY, AWS_SECRET_KEY)
conn_as = AutoScaleConnection(AWS_ACCESS_KEY, AWS_SECRET_KEY)
#=================Create a Load Balancer=============================================
#For a complete list of options see http://boto.cloudhackers.com/ref/ec2.html#module-boto.ec2.elb.healthcheck
hc = HealthCheck('healthCheck',
interval=elastic_load_balancer['interval'],
target=elastic_load_balancer['health_check_target'],
timeout=elastic_load_balancer['timeout'])
#For a complete list of options see http://boto.cloudhackers.com/ref/ec2.html#boto.ec2.elb.ELBConnection.create_load_balancer
lb = conn_elb.create_load_balancer(
elastic_load_balancer['name'], zoneStrings,
elastic_load_balancer['connection_forwarding'])
lb.configure_health_check(hc)
def setup(CONF):
global out
lookup_tbl = {
'name': CONF['NAME'],
}
conn = AutoScaleConnection()
out['conn'] = conn
# Launch Configurations
LC = CONF['LC']
LC['name'] = LC['name'] % lookup_tbl
lc = LaunchConfiguration(**LC)
conn.create_launch_configuration(lc)
out['lc'] = lc
# Auto Scaling Group
ASG = CONF['ASG']
ASG['group_name'] = ASG['group_name'] % lookup_tbl
ASG['launch_config'] = lc
groups = conn.get_all_groups(names=[ASG['group_name']])
if (len(groups) > 0):
# update
asg = groups[0]
for k in ASG :
# asg not iterable, try-except to make sure asg[k] exists
try: asg.__getattribute__(k)
except: continue
asg.__setattr__(k, ASG[k])
asg.launch_config_name = LC['name']
asg.update()
out['asg'] = asg
else:
#create
asg = AutoScalingGroup(**ASG)
conn.create_auto_scaling_group(asg)
# ASG Tags
ASG_TAGS = CONF['ASG_TAGS']
for i in ASG_TAGS:
if 'propagate_at_launch' not in i:
i['propagate_at_launch'] = True
i['key'] = i['key'] % lookup_tbl
i['value'] = i['value'] % lookup_tbl
tags = [
Tag(**dict(x.items() + [('resource_id', ASG['group_name'])])) for x in ASG_TAGS
]
conn.create_or_update_tags(tags)
# Triggers (Scaling Policy / Cloudwatch Alarm)
conn_cw = connect_to_region(CONF['REGION'])
TRIGGERS = CONF['TRIGGERS']
for T in TRIGGERS:
T['policy']['name'] = T['policy']['name'] % lookup_tbl
T['policy']['as_name'] = ASG['group_name']
T['alarm']['dimensions'] = {'AutoScalingGroupName': ASG['group_name']}
T['alarm']['alarm_actions'] = None
if 'name' in T['alarm']:
T['alarm']['name'] = T['alarm']['name'] % lookup_tbl
else:
T['alarm']['name'] = T['policy']['name']
# Policies are safely overwritten, so not checked for existence
conn.create_scaling_policy(ScalingPolicy(**T['policy']))
policy = conn.get_all_policies(as_group=ASG['group_name'], policy_names=[T['policy']['name']])[0]
T['alarm']['alarm_actions'] = [policy.policy_arn]
hits = conn_cw.describe_alarms(alarm_names=[T['alarm']['name']])
conn_cw.create_alarm(MetricAlarm(**T['alarm']))
