Select Git revision
GrabbableBox.uasset
-
David Gilbert authoredDavid Gilbert authored
demo4-scale-out-lb-in-aws.py 20.92 KiB
import configparser
from os.path import expanduser
# import libcloud.security
import time
from libcloud.compute.base import NodeImage
from libcloud.compute.base import NodeState
from libcloud.compute.providers import get_driver as compute_get_driver
from libcloud.compute.types import Provider as compute_Provider
from libcloud.loadbalancer.base import Member, Algorithm
from libcloud.loadbalancer.types import Provider as loadbalancer_Provider
from libcloud.loadbalancer.providers import get_driver as loadbalancer_get_driver
home = expanduser("~")
# requirements:
# services: EC2, ELB
# resources: 2 instances, (1 keypair, 2 security groups),
# 1 (Classic) Elastic Load Balancer, expensive! delete it after you used it!
# The image to look for and use for the started instance
# aws ec2 describe-images --owner amazon | grep ubuntu | grep jammy | grep hvm | grep ssd |grep amd64 | grep -v minimal | grep -v pro | grep -v testing | grep -v k8s | grep "Name"
ubuntu_image_name = 'ubuntu/images/hvm-ssd/ubuntu-jammy-22.04-amd64-server-20240319'
# ubuntu_image_name = 'ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20210128'
# The public key to be used for SSH connection, please make sure, that you have the corresponding private key
# use existing vockey in AWS Lab env from vocareum, enables login directly
# from the lab's terminal:
#
# ssh -i ~/.ssh/labuser.pem ubuntu@<public-ip>
keypair_name = "vockey"
# keypair_name = 'srieger-pub'
pub_key_file = home + '/.ssh/id_rsa.pub'
# id_rsa.pub should look like this (standard sshd pubkey format):
# ssh-rsa AAAAB3NzaC1yc2EAAAABIwAAAQEAw+J...F3w2mleybgT1w== user@HOSTNAME
# flavor_name = 't2.nano'
# t2.nano only has 512 MB RAM, pip install will cause out of memory (OOM), install-aws.sh cloud-init script uses swap
# to circumvent this issue, but t2.micro is also cheap and has 1 GB RAM which is sufficient for faafo example
flavor_name = 't2.micro'
# default region
# region_name = 'eu-central-1'
# region_name = 'ap-south-1'
# AWS Academy Labs only allow us-east-1 and us-west-1 see our AWS Academy Lab Guide, https://awsacademy.instructure.com/login/
region_name = 'us-east-1'
# region_name = 'us-west-1'
# starting instances in AWS Academy takes significantly longer compared to paid AWS accounts, allow ~ >2 minutes timeout
timeout = 600
def main():
###########################################################################
#
# get credentials
#
###########################################################################
# see AWS Academy Lab for Account Details
# read credentials from file
config = configparser.ConfigParser()
config.read_file(open(home + '/.aws/credentials'))
aws_access_key_id = config['default']['aws_access_key_id']
aws_secret_access_key = config['default']['aws_secret_access_key']
aws_session_token = config['default']['aws_session_token']
# hard coded AWS credentials using vars
# aws_access_key_id = "ASIAX..."
# aws_secret_access_key = "WLxxXK+..."
# aws_session_token = "FwoGZXIvYXdzEMb//////////wEaDE5rX.......0SleZ+L75I9iEri9LA4hovWul8HvexhCBK8.......................Ae/T+VkUbcQRtJEDwg+gYCABuk0JlSj5Wk7YA65r3BSNJXZFpkhbek6VBjvE/cEt5fKZEhENcdFxjAcAJLd6bOWi/oGXU5e3PX3mcXgm0oJpz6h3wqD1LvSDtw5GDwn0BHiF1Mu.......................cm/VukK5F"
###########################################################################
#
# create connection
#
###########################################################################
provider = compute_get_driver(compute_Provider.EC2)
conn = provider(key=aws_access_key_id,
secret=aws_secret_access_key,
token=aws_session_token,
region=region_name)
###########################################################################
#
# get image, flavor, network for instance creation
#
###########################################################################
# print("Search for AMI...")
# image = conn.list_images(ex_filters={"name": ubuntu_image_name})[0]
# print("Using image: %s" % image)
# print("Fetching images (AMI) list from AWS region. This will take a lot of seconds (AWS has a very long list of "
# "supported operating systems and versions)... please be patient...")
# image = ''
# for img in images:
# # if img.name == ubuntu_image_name:
# if img.extra['owner_alias'] == 'amazon':
# print(img)
# if img.id == ubuntu_image_name:
# image = img
# select image directly to save time, as retrieving the image list takes several minutes now,
# need to change ami id here if updated or for other regions, id is working for course in
# summer term 2022, in region: us-east-1 and pointing to ubuntu 18.04 used in the instance wizard,
# to update AMI id use the create instance wizard and copy amd64 image id for ubuntu 18.04 in the
# desired region
#
print("Selecting AMI...")
# us-east-1 examples as of 23.05.2024:
#
# https://cloud-images.ubuntu.com/locator/ec2/
#
# Canonical, Ubuntu, 22.04 LTS, amd64 jammy image build on 2022-04-20
image_id = "ami-012485deee5681dc0"
#
image = conn.list_images(ex_image_ids=[image_id])[0]
print("Using image: %s" % image)
flavors = conn.list_sizes()
flavor = [s for s in flavors if s.id == flavor_name][0]
print(flavor)
# networks = conn.ex_list_networks()
# network = ''
# for net in networks:
# if net.name == project_network:
# network = net
###########################################################################
#
# create keypair dependency
#
###########################################################################
print('Checking for existing SSH key pair...')
keypair_exists = False
for keypair in conn.list_key_pairs():
if keypair.name == keypair_name:
keypair_exists = True
if keypair_exists:
print('Keypair ' + keypair_name + ' already exists. Skipping import.')
else:
print('adding keypair...')
conn.import_key_pair_from_file(keypair_name, pub_key_file)
for keypair in conn.list_key_pairs():
print(keypair)
###########################################################################
#
# clean up resources from previous demos
#
###########################################################################
# destroy running demo instances
for instance in conn.list_nodes():
if instance.name in ['all-in-one', 'app-worker-1', 'app-worker-2', 'app-worker-3', 'app-controller',
'app-services', 'app-api-1', 'app-api-2']:
if instance.state is not NodeState.TERMINATED:
print('Destroying Instance: %s' % instance.name)
conn.destroy_node(instance)
# wait until all nodes are destroyed to be able to remove dependent security groups
nodes_still_running = True
while nodes_still_running:
nodes_still_running = False
time.sleep(3)
instances = conn.list_nodes()
for instance in instances:
# if we see any demo instances still running continue to wait for them to stop
if instance.name in ['all-in-one', 'app-worker-1', 'app-worker-2', 'app-worker-3', 'app-controller',
'app-services', 'app-api-1', 'app-api-2']:
if instance.state is not NodeState.TERMINATED:
nodes_still_running = True
if nodes_still_running is True:
print('There are still instances running, waiting for them to be destroyed...')
else:
print('No instances running')
# delete security groups, respecting dependencies (hence deleting 'control' and 'services' first)
for group in conn.ex_list_security_groups():
if group in ['control', 'services']:
print('Deleting security group: %s' % group)
conn.ex_delete_security_group(group)
# now we can delete security groups 'api' and 'worker', as 'control' and 'api' depended on them, otherwise AWS will
# throw DependencyViolation: resource has a dependent object
for group in conn.ex_list_security_groups():
if group in ['api', 'worker']:
print('Deleting security group: %s' % group)
conn.ex_delete_security_group(group)
###########################################################################
#
# create security group dependency
#
###########################################################################
def get_security_group(connection, security_group_name):
"""A helper function to check if security group already exists"""
print('Checking for existing ' + security_group_name + ' security group...')
for security_grp in connection.ex_list_security_groups():
if security_grp == security_group_name:
print('Security Group ' + security_group_name + ' already exists. Skipping creation.')
return security_grp['group_id']
return False
if not get_security_group(conn, "api"):
api_security_group_result = conn.ex_create_security_group('api', 'for API services only')
api_security_group_id = api_security_group_result['group_id']
conn.ex_authorize_security_group_ingress(api_security_group_id, 22, 22, cidr_ips=['0.0.0.0/0'],
protocol='tcp')
conn.ex_authorize_security_group_ingress(api_security_group_id, 80, 80, cidr_ips=['0.0.0.0/0'],
protocol='tcp')
else:
api_security_group_id = get_security_group(conn, "api")
if not get_security_group(conn, "worker"):
worker_security_group_result = conn.ex_create_security_group('worker', 'for services that run on a worker node')
worker_security_group_id = worker_security_group_result['group_id']
conn.ex_authorize_security_group_ingress(worker_security_group_id, 22, 22, cidr_ips=['0.0.0.0/0'],
protocol='tcp')
else:
worker_security_group_id = get_security_group(conn, "worker")
if not get_security_group(conn, "control"):
controller_security_group_result = conn.ex_create_security_group('control',
'for services that run on a control node')
controller_security_group_id = controller_security_group_result['group_id']
conn.ex_authorize_security_group_ingress(controller_security_group_id, 22, 22, cidr_ips=['0.0.0.0/0'],
protocol='tcp')
conn.ex_authorize_security_group_ingress(controller_security_group_id, 80, 80, cidr_ips=['0.0.0.0/0'],
protocol='tcp')
conn.ex_authorize_security_group_ingress(controller_security_group_id, 5672, 5672,
group_pairs=[{'group_id': worker_security_group_id}], protocol='tcp')
else:
controller_security_group_id = get_security_group(conn, "control")
if not get_security_group(conn, "services"):
services_security_group_result = conn.ex_create_security_group('services', 'for DB and AMQP services only')
services_security_group_id = services_security_group_result['group_id']
conn.ex_authorize_security_group_ingress(services_security_group_id, 22, 22, cidr_ips=['0.0.0.0/0'],
protocol='tcp')
# conn.ex_authorize_security_group_ingress(services_security_group_id, 3306, 3306, cidr_ips=['0.0.0.0/0'],
# group_pairs=[{'group_id': api_security_group_id}], protocol='tcp')
conn.ex_authorize_security_group_ingress(services_security_group_id, 3306, 3306,
group_pairs=[{'group_id': api_security_group_id}], protocol='tcp')
conn.ex_authorize_security_group_ingress(services_security_group_id, 5672, 5672,
group_pairs=[{'group_id': worker_security_group_id}], protocol='tcp')
conn.ex_authorize_security_group_ingress(services_security_group_id, 5672, 5672,
group_pairs=[{'group_id': api_security_group_id}], protocol='tcp')
else:
services_security_group_id = get_security_group(conn, "services")
for security_group in conn.ex_list_security_groups():
print(security_group)
# get availability zones
az = conn.list_locations()
print(az)
###########################################################################
#
# create app-services instance (database & messaging) (Amazon AWS EC2)
#
###########################################################################
# https://git.openstack.org/cgit/openstack/faafo/plain/contrib/install-aws.sh
# is currently broken, hence the "rabbitctl" lines were added in the example
# below, see also https://bugs.launchpad.net/faafo/+bug/1679710
#
# Thanks to Stefan Friedmann for finding this fix ;)
userdata_service = '''#!/usr/bin/env bash
curl -L -s https://gogs.informatik.hs-fulda.de/srieger/cloud-computing-msc-ai-examples/raw/master/faafo/contrib/install-aws.sh | bash -s -- \
-i database -i messaging
rabbitmqctl add_user faafo guest
rabbitmqctl set_user_tags faafo administrator
rabbitmqctl set_permissions -p / faafo ".*" ".*" ".*"
'''
print('Starting new app-services instance and wait until it is running...')
instance_services = conn.create_node(location=az[0],
name='app-services',
image=image,
size=flavor,
ex_keyname=keypair_name,
ex_userdata=userdata_service,
ex_security_groups=["services"])
instance_services = conn.wait_until_running(nodes=[instance_services], timeout=timeout, ssh_interface='public_ips')
services_ip = instance_services[0][0].private_ips[0]
print(instance_services)
###########################################################################
#
# create app-api instances (Amazon AWS EC2)
#
###########################################################################
userdata_api = '''#!/usr/bin/env bash
curl -L -s https://gogs.informatik.hs-fulda.de/srieger/cloud-computing-msc-ai-examples/raw/master/faafo/contrib/install-aws.sh | bash -s -- \
-i faafo -r api -m 'amqp://faafo:guest@%(services_ip)s:5672/' \
-d 'mysql+pymysql://faafo:password@%(services_ip)s:3306/faafo'
''' % {'services_ip': services_ip}
print('Starting new app-api-1 instance and wait until it is running...')
instance_api_1 = conn.create_node(location=az[0],
name='app-api-1',
image=image,
size=flavor,
ex_keyname=keypair_name,
ex_userdata=userdata_api,
ex_security_groups=["api"])
print('Starting new app-api-2 instance and wait until it is running...')
instance_api_2 = conn.create_node(location=az[1],
name='app-api-2',
image=image,
size=flavor,
ex_keyname=keypair_name,
ex_userdata=userdata_api,
ex_security_groups=["api"])
instance_api_1 = conn.wait_until_running(nodes=[instance_api_1], timeout=timeout, ssh_interface='public_ips')
api_1_ip = instance_api_1[0][0].private_ips[0]
print("app-api-1 public ip: " + instance_api_1[0][1][0])
instance_api_2 = conn.wait_until_running(nodes=[instance_api_2], timeout=timeout, ssh_interface='public_ips')
# currently only api_1_ip is used
api_2_ip = instance_api_2[0][0].private_ips[0]
print("app-api-2 public ip: " + instance_api_2[0][1][0])
###########################################################################
#
# create worker instances (Amazon AWS EC2)
#
###########################################################################
userdata_worker = '''#!/usr/bin/env bash
curl -L -s https://gogs.informatik.hs-fulda.de/srieger/cloud-computing-msc-ai-examples/raw/master/faafo/contrib/install-aws.sh | bash -s -- \
-i faafo -r worker -e 'http://%(api_1_ip)s' -m 'amqp://faafo:guest@%(services_ip)s:5672/'
''' % {'api_1_ip': api_1_ip, 'services_ip': services_ip}
# userdata_api-api-2 = '''#!/usr/bin/env bash
# curl -L -s https://gogs.informatik.hs-fulda.de/srieger/cloud-computing-msc-ai-examples/raw/master/faafo/contrib/install-aws.sh | bash -s -- \
# -i faafo -r worker -e 'http://%(api_2_ip)s' -m 'amqp://faafo:guest@%(services_ip)s:5672/'
# ''' % {'api_2_ip': api_2_ip, 'services_ip': services_ip}
print('Starting new app-worker-1 instance and wait until it is running...')
instance_worker_1 = conn.create_node(location=az[0],
name='app-worker-1',
image=image, size=flavor,
ex_keyname=keypair_name,
ex_userdata=userdata_worker,
ex_security_groups=["worker"])
print('Starting new app-worker-2 instance and wait until it is running...')
instance_worker_2 = conn.create_node(location=az[1],
name='app-worker-2',
image=image, size=flavor,
ex_keyname=keypair_name,
ex_userdata=userdata_worker,
ex_security_groups=["worker"])
# do not start worker 3 initially, can be started using scale-out-add-worker.py demo
# print('Starting new app-worker-3 instance and wait until it is running...')
# instance_worker_3 = conn.create_node(name='app-worker-3',
# image=image, size=flavor,
# networks=[network],
# ex_keyname=keypair_name,
# ex_userdata=userdata_worker,
# ex_security_groups=[worker_security_group])
print(instance_worker_1)
print(instance_worker_2)
# print(instance_worker_3)
###########################################################################
#
# create load balancer (Amazon AWS ELB)
#
###########################################################################
elb_provider = loadbalancer_get_driver(loadbalancer_Provider.ELB)
elb_conn = elb_provider(aws_access_key_id,
aws_secret_access_key,
token=aws_session_token,
region=region_name)
print("Deleting previously created load balancers in: " + str(elb_conn.list_balancers()))
for loadbalancer in elb_conn.list_balancers():
if loadbalancer.name == "lb1":
print("Deleting Load Balancer: " + str(loadbalancer))
elb_conn.destroy_balancer(loadbalancer)
# get suffix (a, b, c, ...) from all availability zones, available in the selected region
all_availability_zones_in_region = []
for az in conn.ex_list_availability_zones():
all_availability_zones_in_region.append(az.name[-1])
# create new load balancer
# example uses "classic" ELB with default HTTP health. monitor, you can see the result in the EC2 console, after
# running this script
new_load_balancer = elb_conn.create_balancer(
name='lb1',
algorithm=Algorithm.ROUND_ROBIN,
port=80,
protocol='http',
members=[],
ex_members_availability_zones=all_availability_zones_in_region)
# attach api instances as members to load balancer
elb_conn.balancer_attach_compute_node(balancer=new_load_balancer, node=instance_api_1[0][0])
elb_conn.balancer_attach_compute_node(balancer=new_load_balancer, node=instance_api_2[0][0])
print("Created load balancer: " + str(new_load_balancer))
# wait for the load balancer to be ready
while new_load_balancer.state != 2:
time.sleep(3)
new_load_balancer = elb_conn.get_balancer(new_load_balancer.id)
print("\n\nYou can see the instances created in EC2 in AWS Console. You'll also find the load balancer under ELB "
"there.\n"
" You can access the faafo application deployed to the loadbalancer at: http://" + new_load_balancer.ip +
" as soon as instances are detected to be deployed and healthy by the load balancer.")
if __name__ == '__main__':
main()