def create_config(bosslet_config):
session = bosslet_config.session
names = bosslet_config.names
config = CloudFormationConfiguration("dynamolambda", bosslet_config)
role = aws.role_arn_lookup(session, "lambda_cache_execution")
config.add_arg(
Arg.String("LambdaCacheExecutionRole", role,
"IAM role for " + names.multi_lambda.lambda_))
config.add_lambda(DYNAMO_LAMBDA_KEY,
names.dynamo_lambda.lambda_,
Ref("LambdaCacheExecutionRole"),
handler="index.handler",
timeout=120,
memory=128,
runtime="nodejs8.10",
reserved_executions=1)
config.add_cloudwatch_rule(TRIGGER_KEY,
name=names.trigger_dynamo_autoscale.cw,
description='Run DynamoDB table autoscaler',
targets=[{
'Arn': Arn(DYNAMO_LAMBDA_KEY),
'Id': names.vault_monitor.lambda_,
}],
schedule='rate(1 minute)',
depends_on=[DYNAMO_LAMBDA_KEY])
config.add_lambda_permission('TriggerPerms',
names.dynamo_lambda.lambda_,
principal='events.amazonaws.com',
source=Arn(TRIGGER_KEY))
return config
def create_config(session, domain, peer_domain):
config = CloudFormationConfiguration('peer', domain)
# Reconfigure the stack name from the default
stack_name = domain + ".to." + peer_domain
config.stack_name = "".join(
[x.capitalize() for x in stack_name.split('.')])
vpc_id = aws.vpc_id_lookup(session, domain)
vpc_subnet = hosts.lookup(domain)
names = AWSNames(domain)
peer_id = aws.vpc_id_lookup(session, peer_domain)
peer_subnet = hosts.lookup(peer_domain)
peer_names = AWSNames(peer_domain)
if session is not None:
if vpc_id is None:
raise Exception("VPC does not exist, exiting...")
if peer_id is None:
raise Exception("Peer VPC does not exist, existing...")
config.add_vpc_peering("Peer", vpc_id, peer_id)
def add_route(key, vpc_id_, rt_name, vpc_subnet_):
rt = aws.rt_lookup(session, vpc_id_, rt_name)
config.add_route_table_route(key, rt, vpc_subnet_, peer=Ref("Peer"))
add_route("PeeringRoute", vpc_id, names.internal, peer_subnet)
add_route("PeeringRoute2", vpc_id, names.internet, peer_subnet)
add_route("PeerPeeringRoute", peer_id, peer_names.internal, vpc_subnet)
add_route("PeerPeeringRoute2", peer_id, peer_names.internet, vpc_subnet)
return config
def delete(bosslet_config):
# NOTE: CloudWatch logs for the DNS Lambda are not deleted
session = bosslet_config.session
domain = bosslet_config.INTERNAL_DOMAIN
names = bosslet_config.names
aws.route53_delete_records(session, domain, names.cachemanager.dns)
config = CloudFormationConfiguration("cachedb", bosslet_config)
config.delete()
def create_config(session, domain, keypair=None):
"""
Create the CloudFormationConfiguration object.
Args:
session: amazon session object
domain (string): domain of the stack being created
keypair: keypair used to by instances being created
Returns: the config for the Cloud Formation stack
"""
names = AWSNames(domain)
config = CloudFormationConfiguration('redis', domain, const.REGION)
vpc_id = config.find_vpc(session)
az_subnets, external_subnets = config.find_all_availability_zones(session)
sgs = aws.sg_lookup_all(session, vpc_id)
# Create the Cache and CacheState Redis Clusters
REDIS_PARAMETERS = {
"maxmemory-policy": "volatile-lru",
"reserved-memory": str(get_scenario(const.REDIS_RESERVED_MEMORY, 0) * 1000000),
"maxmemory-samples": "5", # ~ 5 - 10
}
config.add_redis_replication("Cache",
names.cache,
az_subnets,
[sgs[names.internal]],
type_=const.REDIS_CACHE_TYPE,
version="3.2.4",
clusters=const.REDIS_CLUSTER_SIZE,
parameters=REDIS_PARAMETERS)
config.add_redis_replication("CacheState",
names.cache_state,
az_subnets,
[sgs[names.internal]],
type_=const.REDIS_TYPE,
version="3.2.4",
clusters=const.REDIS_CLUSTER_SIZE)
# This one may not be created depending on the scenario type.
if get_scenario(const.REDIS_SESSION_TYPE, None) is not None:
config.add_redis_replication("CacheSession",
names.cache_session,
az_subnets,
[sgs[names.internal]],
type_=const.REDIS_SESSION_TYPE,
version="3.2.4",
clusters=1)
return config
def update_migrate(bosslet_config, config):
migration_progress = constants.repo_path("cloud_formation", "configs", "migrations", config, "progress")
if not os.path.exists(migration_progress):
console.info("No migrations to apply")
return
with open(migration_progress, "r") as fh:
cur_ver = int(fh.read())
new_ver = CloudFormationConfiguration(config, bosslet_config).existing_version()
migrations = MigrationManager(config, cur_ver, new_ver)
if not migrations.has_migrations:
console.info("No migrations to apply")
os.remove(migration_progress)
return
def callback(migration_file):
with open(migration_progress, 'w') as fh:
fh.write(str(migration_file.stop))
migrations.add_callback(post=callback)
migrations.post_update(bosslet_config)
os.remove(migration_progress)
def delete(bosslet_config):
session = bosslet_config.session
domain = bosslet_config.INTERNAL_DOMAIN
names = bosslet_config.names
if not console.confirm(
"All data will be lost. Are you sure you want to proceed?"):
raise BossManageCanceled()
aws.route53_delete_records(session, domain, names.endpoint.dns)
# Other configs may define SQS queues and we shouldn't delete them
aws.sqs_delete_all(
session, domain) # !!! TODO FIX this so it doesn't bork the stack
aws.policy_delete_all(session, domain, '/ingest/')
config = CloudFormationConfiguration('api', bosslet_config)
config.delete()
def create_config(bosslet_config):
names = bosslet_config.names
session = bosslet_config.session
config = CloudFormationConfiguration('redis', bosslet_config)
vpc_id = config.find_vpc()
internal_subnets, external_subnets = config.find_all_subnets()
sgs = aws.sg_lookup_all(session, vpc_id)
# Create the Cache and CacheState Redis Clusters
REDIS_PARAMETERS = {
"maxmemory-policy": "volatile-lru",
"reserved-memory-percent": str(const.REDIS_RESERVED_MEMORY_PERCENT),
"maxmemory-samples": "5", # ~ 5 - 10
}
config.add_redis_replication("Cache",
names.cache.redis,
internal_subnets,
[sgs[names.internal.sg]],
type_=const.REDIS_CACHE_TYPE,
version="3.2.4",
clusters=const.REDIS_CLUSTER_SIZE,
parameters=REDIS_PARAMETERS)
config.add_redis_replication("CacheState",
names.cache_state.redis,
internal_subnets,
[sgs[names.internal.sg]],
type_=const.REDIS_TYPE,
version="3.2.4",
clusters=const.REDIS_CLUSTER_SIZE)
# This one may not be created depending on the scenario type.
if const.REDIS_SESSION_TYPE is not None:
config.add_redis_replication("CacheSession",
names.cache_session.redis,
internal_subnets,
[sgs[names.internal.sg]],
type_=const.REDIS_SESSION_TYPE,
version="3.2.4",
clusters=1)
return config
def create_config(session, domain):
"""
Create the CloudFormationConfiguration object.
Args:
session (Session): amazon session object
domain (str): domain of the stack being created
Returns: the config for the Cloud Formation stack
"""
names = AWSNames(domain)
config = CloudFormationConfiguration("dynamolambda", domain, const.REGION)
role = aws.role_arn_lookup(session, "lambda_cache_execution")
config.add_arg(
Arg.String("LambdaCacheExecutionRole", role,
"IAM role for multilambda." + domain))
lambda_bucket = aws.get_lambda_s3_bucket(session)
lambda_key = generate_lambda_key(domain)
config.add_lambda(DYNAMO_LAMBDA_KEY,
names.dynamo_lambda,
Ref("LambdaCacheExecutionRole"),
s3=(aws.get_lambda_s3_bucket(session), lambda_key,
"index.handler"),
timeout=120,
memory=128,
runtime="nodejs6.10")
config.add_cloudwatch_rule(TRIGGER_KEY,
name=names.trigger_dynamo_autoscale,
description='Run DynamoDB table autoscaler',
targets=[{
'Arn': Arn(DYNAMO_LAMBDA_KEY),
'Id': names.vault_monitor,
}],
schedule='rate(1 minute)',
depends_on=[DYNAMO_LAMBDA_KEY])
config.add_lambda_permission('TriggerPerms',
names.dynamo_lambda,
principal='events.amazonaws.com',
source=Arn(TRIGGER_KEY))
return config
def create_config(session, domain):
"""
Create the CloudFormationConfiguration object.
Args:
session (Session): amazon session object
domain (str): domain of the stack being created
Returns: the config for the Cloud Formation stack
"""
names = AWSNames(domain)
config = CloudFormationConfiguration("dynamolambda", domain, const.REGION)
role = aws.role_arn_lookup(session, "lambda_cache_execution")
config.add_arg(Arg.String("LambdaCacheExecutionRole", role,
"IAM role for multilambda." + domain))
lambda_bucket = aws.get_lambda_s3_bucket(session)
lambda_key = generate_lambda_key(domain)
config.add_lambda(DYNAMO_LAMBDA_KEY,
names.dynamo_lambda,
Ref("LambdaCacheExecutionRole"),
s3=(aws.get_lambda_s3_bucket(session),
lambda_key,
"index.handler"),
timeout=120,
memory=128,
runtime="nodejs6.10",
reserved_executions=1)
config.add_cloudwatch_rule(TRIGGER_KEY,
name=names.trigger_dynamo_autoscale,
description='Run DynamoDB table autoscaler',
targets=[
{
'Arn': Arn(DYNAMO_LAMBDA_KEY),
'Id': names.vault_monitor,
}
],
schedule='rate(1 minute)',
depends_on=[DYNAMO_LAMBDA_KEY])
config.add_lambda_permission('TriggerPerms',
names.dynamo_lambda,
principal='events.amazonaws.com',
source=Arn(TRIGGER_KEY))
return config
def delete(bosslet_config):
# NOTE: CloudWatch logs for the DNS Lambda are not deleted
if not console.confirm(
"All data will be lost. Are you sure you want to proceed?"):
raise BossManageCanceled()
session = bosslet_config.session
domain = bosslet_config.INTERNAL_DOMAIN
names = bosslet_config.names
aws.route53_delete_records(session, domain, names.auth.dns)
aws.route53_delete_records(session, domain, names.vault.dns)
aws.sns_unsubscribe_all(bosslet_config, names.dns.sns)
config = CloudFormationConfiguration('core', bosslet_config)
if config.existing_version(
) == 1: # Deleting a stack that has not been updated
aws.route53_delete_records(session, domain, 'consul.' + domain)
config.delete()
def delete(session, domain):
names = AWSNames(domain)
# DP TODO: delete activities
CloudFormationConfiguration('activities', domain).delete(session)
sfn.delete(session, names.delete_cuboid)
sfn.delete(session, names.delete_experiment)
sfn.delete(session, names.delete_coord_frame)
sfn.delete(session, names.delete_collection)
sfn.delete(session, names.query_deletes)
sfn.delete(session, names.ingest_queue_populate)
sfn.delete(session, names.ingest_queue_upload)
sfn.delete(session, names.resolution_hierarchy)
def call_config(session, domain, config, func_name):
"""Import 'configs.<config>' and then call the requested function with
<session> and <domain>.
"""
module = importlib.import_module("configs." + config)
if func_name in module.__dict__:
module.__dict__[func_name](session, domain)
elif func_name == 'delete':
CloudFormationConfiguration(config, domain).delete(session)
else:
print("Configuration '{}' doesn't implement function '{}'".format(
config, func_name))
def create_config(session, domain, peer_domain):
config = CloudFormationConfiguration('peer', domain)
# Reconfigure the stack name from the default
stack_name = domain + ".to." + peer_domain
config.stack_name = "".join([x.capitalize() for x in stack_name.split('.')])
vpc_id = aws.vpc_id_lookup(session, domain)
vpc_subnet = hosts.lookup(domain)
names = AWSNames(domain)
peer_id = aws.vpc_id_lookup(session, peer_domain)
peer_subnet = hosts.lookup(peer_domain)
peer_names = AWSNames(peer_domain)
if session is not None:
if vpc_id is None:
raise Exception("VPC does not exist, exiting...")
if peer_id is None:
raise Exception("Peer VPC does not exist, existing...")
config.add_vpc_peering("Peer",
vpc_id,
peer_id)
def add_route(key, vpc_id_, rt_name, vpc_subnet_):
rt = aws.rt_lookup(session, vpc_id_, rt_name)
config.add_route_table_route(key, rt, vpc_subnet_, peer = Ref("Peer"))
add_route("PeeringRoute", vpc_id, names.internal, peer_subnet)
add_route("PeeringRoute2", vpc_id, names.internet, peer_subnet)
add_route("PeerPeeringRoute", peer_id, peer_names.internal, vpc_subnet)
add_route("PeerPeeringRoute2", peer_id, peer_names.internet, vpc_subnet)
return config
def create_config(session, domain, keypair=None):
"""
Create the CloudFormationConfiguration object.
Args:
session: amazon session object
domain (string): domain of the stack being created
keypair: keypair used to by instances being created
Returns: the config for the Cloud Formation stack
"""
names = AWSNames(domain)
config = CloudFormationConfiguration('redis', domain, const.REGION)
vpc_id = config.find_vpc(session)
az_subnets, external_subnets = config.find_all_availability_zones(session)
sgs = aws.sg_lookup_all(session, vpc_id)
# Create the Cache and CacheState Redis Clusters
REDIS_PARAMETERS = {
"maxmemory-policy": "volatile-lru",
"reserved-memory": str(get_scenario(const.REDIS_RESERVED_MEMORY, 0) * 1000000),
"maxmemory-samples": "5", # ~ 5 - 10
}
config.add_redis_replication("Cache",
names.cache,
az_subnets,
[sgs[names.internal]],
type_=const.REDIS_CACHE_TYPE,
version="3.2.4",
clusters=const.REDIS_CLUSTER_SIZE,
parameters=REDIS_PARAMETERS)
config.add_redis_replication("CacheState",
names.cache_state,
az_subnets,
[sgs[names.internal]],
type_=const.REDIS_TYPE,
version="3.2.4",
clusters=const.REDIS_CLUSTER_SIZE)
return config
def create_config(bosslet_config):
"""Create the CloudFormationConfiguration object."""
config = CloudFormationConfiguration('copycuboid', bosslet_config)
names = bosslet_config.names
session = bosslet_config.session
role = aws.role_arn_lookup(session, "lambda_cache_execution")
config.add_arg(
Arg.String(
"LambdaCacheExecutionRole", role,
"IAM role for multilambda." + bosslet_config.INTERNAL_DOMAIN))
config.add_sqs_queue(names.copy_cuboid_dlq.sqs, names.copy_cuboid_dlq.sqs,
30, 20160)
config.add_lambda("CopyCuboidLambda",
names.copy_cuboid_lambda.lambda_,
Ref("LambdaCacheExecutionRole"),
handler="copy_cuboid_lambda.handler",
timeout=60,
memory=128,
dlq=Arn(names.copy_cuboid_dlq.sqs))
return config
def call_configs(bosslet_config, configs, func_name):
"""Import 'configs.<config>' and then call the requested function with
<session> and <bosslet>.
"""
modules = [(config, importlib.import_module("configs." + config)) for config in configs]
if func_name != 'generate':
modules = build_dependency_graph(func_name, bosslet_config, modules)
print("Execution Order:")
for config, module in modules:
print("\t{}".format(config))
with console.status_line(spin=True, print_status=True) as status:
for config, module in modules:
print()
status('Working on {}'.format(config))
if func_name in module.__dict__:
module.__dict__[func_name](bosslet_config)
elif func_name == 'delete':
CloudFormationConfiguration(config, bosslet_config).delete()
else:
print("Configuration '{}' doesn't implement function '{}', skipping".format(config, func_name))
def create_config(session, domain):
"""Create the CloudFormationConfiguration object."""
config = CloudFormationConfiguration('copycuboid', domain, const.REGION)
names = AWSNames(domain)
role = aws.role_arn_lookup(session, "lambda_cache_execution")
config.add_arg(Arg.String(
"LambdaCacheExecutionRole", role,
"IAM role for multilambda." + domain))
config.add_sqs_queue(names.copy_cuboid_dlq, names.copy_cuboid_dlq, 30, 20160)
config.add_lambda("CopyCuboidLambda",
names.copy_cuboid_lambda,
Ref("LambdaCacheExecutionRole"),
s3=(aws.get_lambda_s3_bucket(session),
"multilambda.{}.zip".format(domain),
"copy_cuboid_lambda.handler"),
timeout=60,
memory=128,
runtime='python3.6',
dlq=Arn(names.copy_cuboid_dlq))
return config
def create_config(session, domain, keypair=None, user_data=None):
"""
Create the CloudFormationConfiguration object.
Args:
session: amazon session object
domain: domain of the stack being created
keypair: keypair used to by instances being created
user_data (UserData): information used by the endpoint instance and vault. Data will be run through the CloudFormation Fn::Join template intrinsic function so other template intrinsic functions used in the user_data will be parsed and executed.
Returns: the config for the Cloud Formation stack
"""
# Prepare user data for parsing by CloudFormation.
if user_data is not None:
parsed_user_data = { "Fn::Join" : ["", user_data.format_for_cloudformation()]}
else:
parsed_user_data = user_data
names = AWSNames(domain)
config = CloudFormationConfiguration("cachedb", domain, const.REGION)
vpc_id = config.find_vpc(session)
# Create several subnets for all the lambdas to use.
lambda_azs = aws.azs_lookup(session, lambda_compatible_only=True)
internal_route_table_id = aws.rt_lookup(session, vpc_id, names.internal)
print("AZs for lambda: " + str(lambda_azs))
lambda_subnets = []
for i in range(const.LAMBDA_SUBNETS):
key = 'LambdaSubnet{}'.format(i)
lambda_subnets.append(Ref(key))
config.add_subnet(key, names.subnet('lambda{}'.format(i)), az=lambda_azs[i % len(lambda_azs)][0])
config.add_route_table_association(key + "RTA",
internal_route_table_id,
Ref(key))
# Lookup the External Subnet, Internal Security Group IDs that are
# needed by other resources
internal_subnet_id = aws.subnet_id_lookup(session, names.subnet("internal"))
config.add_arg(Arg.Subnet("InternalSubnet",
internal_subnet_id,
"ID of Internal Subnet to create resources in"))
internal_sg_id = aws.sg_lookup(session, vpc_id, names.internal)
config.add_arg(Arg.SecurityGroup("InternalSecurityGroup",
internal_sg_id,
"ID of internal Security Group"))
role = aws.role_arn_lookup(session, "lambda_cache_execution")
config.add_arg(Arg.String("LambdaCacheExecutionRole", role,
"IAM role for multilambda." + domain))
cuboid_import_role = aws.role_arn_lookup(session, CUBOID_IMPORT_ROLE)
config.add_arg(Arg.String(CUBOID_IMPORT_ROLE, cuboid_import_role,
"IAM role for cuboidImport." + domain))
config.add_capabilities(['CAPABILITY_IAM'])
# Allow updating S3 index table with cuboid's object key during
# volumetric ingest.
# Example of s3_index_arn form: arn:aws:dynamodb:us-east-1:12345678:table/s3index.*.boss
config.add_iam_policy_to_role(
'S3IndexPutItem{}'.format(domain).replace('.', ''),
get_s3_index_arn(session, domain).replace(domain,'*.') + domain.split('.')[1],
[CUBOID_IMPORT_ROLE], ['dynamodb:PutItem'])
cuboid_bucket_name = names.cuboid_bucket
if not aws.s3_bucket_exists(session, cuboid_bucket_name):
config.add_s3_bucket("cuboidBucket", cuboid_bucket_name)
config.add_s3_bucket_policy(
"cuboidBucketPolicy", cuboid_bucket_name,
['s3:GetObject', 's3:PutObject'],
{ 'AWS': role})
config.append_s3_bucket_policy(
"cuboidBucketPolicy", cuboid_bucket_name,
['s3:PutObject'], { 'AWS': cuboid_import_role})
delete_bucket_name = names.delete_bucket
if not aws.s3_bucket_exists(session, delete_bucket_name):
config.add_s3_bucket("deleteBucket", delete_bucket_name)
config.add_s3_bucket_policy(
"deleteBucketPolicy", delete_bucket_name,
['s3:GetObject', 's3:PutObject'],
{ 'AWS': role})
tile_bucket_name = names.tile_bucket
if not aws.s3_bucket_exists(session, tile_bucket_name):
life_cycle_cfg = get_cf_bucket_life_cycle_rules()
config.add_s3_bucket(
"tileBucket", tile_bucket_name, life_cycle_config=life_cycle_cfg)
config.add_s3_bucket_policy(
"tileBucketPolicy", tile_bucket_name,
['s3:GetObject', 's3:PutObject'],
{ 'AWS': role})
# The ingest bucket is a staging area for cuboids uploaded during volumetric ingest.
creating_ingest_bucket = False
ingest_bucket_name = names.ingest_bucket
if not aws.s3_bucket_exists(session, ingest_bucket_name):
creating_ingest_bucket = True
ing_bucket_life_cycle_cfg = get_cf_bucket_life_cycle_rules()
config.add_s3_bucket("ingestBucket", ingest_bucket_name,
life_cycle_config=ing_bucket_life_cycle_cfg)
config.add_s3_bucket_policy(
"ingestBucketPolicy", ingest_bucket_name,
['s3:GetObject', 's3:PutObject', 's3:PutObjectTagging'],
{ 'AWS': cuboid_import_role})
config.add_ec2_instance("CacheManager",
names.cache_manager,
aws.ami_lookup(session, "cachemanager.boss"),
keypair,
subnet=Ref("InternalSubnet"),
public_ip=False,
type_=const.CACHE_MANAGER_TYPE,
security_groups=[Ref("InternalSecurityGroup")],
user_data=parsed_user_data,
role="cachemanager")
config.add_sqs_queue(
names.ingest_cleanup_dlq, names.ingest_cleanup_dlq, 30, 20160)
config.add_sqs_queue(
names.cuboid_import_dlq, names.cuboid_import_dlq, 30, 20160)
config.add_sqs_policy('cuboidImportDlqPolicy', 'cuboidImportDlqPolicy',
[Ref(names.cuboid_import_dlq)], cuboid_import_role)
lambda_bucket = aws.get_lambda_s3_bucket(session)
config.add_lambda("MultiLambda",
names.multi_lambda,
Ref("LambdaCacheExecutionRole"),
s3=(lambda_bucket,
"multilambda.{}.zip".format(domain),
"lambda_loader.handler"),
timeout=120,
memory=1536,
security_groups=[Ref('InternalSecurityGroup')],
subnets=lambda_subnets,
runtime='python3.6')
config.add_lambda("TileUploadedLambda",
names.tile_uploaded_lambda,
Ref("LambdaCacheExecutionRole"),
s3=(lambda_bucket,
"multilambda.{}.zip".format(domain),
"tile_uploaded_lambda.handler"),
timeout=5,
memory=1024,
runtime='python3.6')
config.add_lambda("TileIngestLambda",
names.tile_ingest_lambda,
Ref("LambdaCacheExecutionRole"),
s3=(lambda_bucket,
"multilambda.{}.zip".format(domain),
"tile_ingest_lambda.handler"),
timeout=30,
memory=1536,
runtime='python3.6')
config.add_lambda("DeleteTileObjsLambda",
names.delete_tile_objs_lambda,
Ref("LambdaCacheExecutionRole"),
s3=(lambda_bucket,
"multilambda.{}.zip".format(domain),
"delete_tile_objs_lambda.handler"),
timeout=90,
memory=128,
runtime='python3.6',
dlq=Arn(names.ingest_cleanup_dlq))
config.add_lambda("DeleteTileEntryLambda",
names.delete_tile_index_entry_lambda,
Ref("LambdaCacheExecutionRole"),
s3=(lambda_bucket,
"multilambda.{}.zip".format(domain),
"delete_tile_index_entry_lambda.handler"),
timeout=90,
memory=128,
runtime='python3.6',
dlq=Arn(names.ingest_cleanup_dlq))
config.add_lambda("CuboidImportLambda",
names.cuboid_import_lambda,
Ref(CUBOID_IMPORT_ROLE),
s3=(lambda_bucket,
"multilambda.{}.zip".format(domain),
"cuboid_import_lambda.handler"),
timeout=90,
memory=128,
runtime='python3.6',
dlq=Arn(names.cuboid_import_dlq))
config.add_lambda("VolumetricIngestLambda",
names.volumetric_ingest_queue_upload_lambda,
Ref("LambdaCacheExecutionRole"),
s3=(lambda_bucket,
"multilambda.{}.zip".format(domain),
"ingest_queue_upload_volumetric_lambda.handler"),
timeout=120,
memory=1024,
runtime='python3.6')
if creating_ingest_bucket:
config.add_lambda_permission(
'ingestBucketInvokeCuboidImportLambda', names.cuboid_import_lambda,
principal='s3.amazonaws.com', source={
'Fn::Join': [':', ['arn', 'aws', 's3', '', '', ingest_bucket_name]]}, #DP TODO: move into constants
depends_on=['ingestBucket', 'CuboidImportLambda']
)
else:
# NOTE: this permission doesn't seem to apply properly when doing a
# CloudFormation update. During testing, I had to manually apply this
# permission before the bucket trigger could be applied in post_init().
# Doing a CloudFormation delete followed by a create did not have a
# problem.
config.add_lambda_permission(
'ingestBucketInvokeCuboidImportLambda', names.cuboid_import_lambda,
principal='s3.amazonaws.com', source={
'Fn::Join': [':', ['arn', 'aws', 's3', '', '', ingest_bucket_name]]},
depends_on='CuboidImportLambda'
)
# Add topic to indicating that the object store has been write locked.
# Now using "production mailing list" instead of separate write lock topic.
#config.add_sns_topic('WriteLock',
# names.write_lock_topic,
# names.write_lock,
# []) # TODO: add subscribers
return config
def delete(bosslet_config):
config = CloudFormationConfiguration('backup', bosslet_config)
config.delete()
def create_config(bosslet_config):
"""Create the CloudFormationConfiguration object."""
config = CloudFormationConfiguration('core', bosslet_config, version="2")
session = bosslet_config.session
keypair = bosslet_config.SSH_KEY
names = bosslet_config.names
config.add_vpc()
# Create the internal and external subnets
config.add_subnet('InternalSubnet', names.internal.subnet)
config.add_subnet('ExternalSubnet', names.external.subnet)
internal_subnets, external_subnets = config.add_all_subnets()
internal_subnets_asg, external_subnets_asg = config.find_all_subnets('asg')
# Create a custom resource to help delete ENIs from lambdas
# DP NOTE: After deleting a lambda the ENIs may stick around for while, causing the stack delete to fail
# See https://stackoverflow.com/a/41310289
config.add_arg(Arg.String('StackName', config.stack_name))
config.add_custom_resource('DeleteENI',
'DeleteENI',
Arn('DeleteENILambda'),
StackName=Ref('StackName'))
config.add_lambda(
"DeleteENILambda",
names.delete_eni.lambda_,
aws.role_arn_lookup(session, 'DeleteENI'),
const.DELETE_ENI_LAMBDA,
handler="index.handler",
timeout=
180, # 3 minutes, so that there is enough time to wait for the ENI detach to complete
runtime='python3.6'
) # If the lambda times out CF will retry a couple of times
user_data = const.BASTION_USER_DATA.format(bosslet_config.NETWORK)
config.add_ec2_instance("Bastion",
names.bastion.dns,
aws.ami_lookup(bosslet_config, const.BASTION_AMI),
keypair,
subnet=Ref("ExternalSubnet"),
public_ip=True,
user_data=user_data,
security_groups=[
Ref("InternalSecurityGroup"),
Ref("BastionSecurityGroup")
],
depends_on="AttachInternetGateway")
vault_role = aws.role_arn_lookup(session, 'apl-vault')
vault_actions = ['kms:Encrypt', 'kms:Decrypt', 'kms:DescribeKey']
config.add_kms_key("VaultKey", names.vault.key, vault_role, vault_actions)
config.add_dynamo_table("VaultTable",
names.vault.ddb,
attributes=[('Path', 'S'), ('Key', 'S')],
key_schema=[('Path', 'HASH'), ('Key', 'RANGE')],
throughput=(5, 5))
user_data = UserData()
user_data["system"]["fqdn"] = names.vault.dns
user_data["system"]["type"] = "vault"
user_data["vault"]["kms_key"] = str(Ref("VaultKey"))
user_data["vault"]["ddb_table"] = names.vault.ddb
parsed_user_data = {
"Fn::Join": ["", user_data.format_for_cloudformation()]
}
config.add_autoscale_group("Vault",
names.vault.dns,
aws.ami_lookup(bosslet_config, names.vault.ami),
keypair,
subnets=internal_subnets_asg,
type_=const.VAULT_TYPE,
security_groups=[Ref("InternalSecurityGroup")],
user_data=parsed_user_data,
min=const.VAULT_CLUSTER_SIZE,
max=const.VAULT_CLUSTER_SIZE,
notifications=Ref("DNSSNS"),
role=aws.instance_profile_arn_lookup(
session, 'apl-vault'),
depends_on=[
"VaultKey", "VaultTable", "DNSLambda",
"DNSSNS", "DNSLambdaExecute"
])
user_data = UserData()
user_data["system"]["fqdn"] = names.auth.dns
user_data["system"]["type"] = "auth"
deps = [
"AuthSecurityGroup", "AttachInternetGateway", "DNSLambda", "DNSSNS",
"DNSLambdaExecute"
]
# Problem: If development scenario uses a local DB. If the auth server crashes
# and is auto restarted by the autoscale group then the new auth server
# will not have any of the previous configuration, because the old DB
# was lost. Using an RDS for development fixes this at the cost of having
# the core config taking longer to launch.
USE_DB = bosslet_config.AUTH_RDS
if USE_DB:
deps.append("AuthDB")
user_data["aws"][
"db"] = "keycloak" # flag for init script for which config to use
cert = aws.cert_arn_lookup(session, names.public_dns('auth'))
create_asg_elb(config,
"Auth",
names.auth.dns,
aws.ami_lookup(bosslet_config, names.auth.ami),
keypair,
str(user_data),
const.AUTH_CLUSTER_SIZE,
internal_subnets_asg,
external_subnets_asg, [("443", "8080", "HTTPS", cert)],
"HTTP:8080/index.html",
sgs=[Ref("AuthSecurityGroup")],
type_=const.AUTH_TYPE,
depends_on=deps)
config.add_public_dns('AuthLoadBalancer', names.public_dns('auth'))
if USE_DB:
config.add_rds_db("AuthDB",
names.auth_db.rds,
"3306",
"keycloak",
"keycloak",
"keycloak",
internal_subnets,
type_="db.t2.micro",
security_groups=[Ref("InternalSecurityGroup")])
config.add_lambda("DNSLambda",
names.dns.lambda_,
aws.role_arn_lookup(session, 'UpdateRoute53'),
const.DNS_LAMBDA,
handler="index.handler",
timeout=10,
depends_on="DNSZone")
config.add_lambda_permission("DNSLambdaExecute", Ref("DNSLambda"))
config.add_sns_topic("DNSSNS", names.dns.sns, names.dns.sns,
[("lambda", Arn("DNSLambda"))])
config.add_security_group("InternalSecurityGroup", names.internal.sg,
[("-1", "-1", "-1", bosslet_config.NETWORK)])
# Allow SSH access to bastion from anywhere
incoming_subnet = bosslet_config.SSH_INBOUND
config.add_security_group("BastionSecurityGroup", names.ssh.sg,
[("tcp", "22", "22", incoming_subnet)])
incoming_subnet = bosslet_config.HTTPS_INBOUND
boss_subnet = {
"Fn::Join": ["/", [Ref("NATIP"), "32"]]
} # Allow requests from the endpoint via the NAT gateway
# Needed in case HTTPS_INBOUND doesn't include the gateway's IP
config.add_security_group(
"AuthSecurityGroup",
#names.https.sg, DP XXX: hack until we can get production updated correctly
names.auth.sg,
[("tcp", "443", "443", incoming_subnet),
("tcp", "443", "443", boss_subnet)])
# Create the internal route table to route traffic to the NAT Bastion
all_internal_subnets = internal_subnets.copy()
all_internal_subnets.append(Ref("InternalSubnet"))
config.add_route_table("InternalRouteTable",
names.internal.rt,
subnets=all_internal_subnets)
config.add_route_table_route("InternalNatRoute",
Ref("InternalRouteTable"),
nat=Ref("NAT"),
depends_on="NAT")
# Create the internet gateway and internet router
all_external_subnets = external_subnets.copy()
all_external_subnets.append(Ref("ExternalSubnet"))
config.add_route_table("InternetRouteTable",
names.internet.rt,
subnets=all_external_subnets)
config.add_route_table_route("InternetRoute",
Ref("InternetRouteTable"),
gateway=Ref("InternetGateway"),
depends_on="AttachInternetGateway")
config.add_internet_gateway("InternetGateway", names.internet.gw)
config.add_endpoint("S3Endpoint", "s3",
[Ref("InternalRouteTable"),
Ref('InternetRouteTable')])
config.add_endpoint("DynamoDBEndpoint", "dynamodb",
[Ref("InternalRouteTable"),
Ref('InternetRouteTable')])
config.add_nat("NAT",
Ref("ExternalSubnet"),
depends_on="AttachInternetGateway")
return config
def delete(session, domain):
names = AWSNames(domain)
CloudFormationConfiguration('redis', domain).delete(session)
def create_config(bosslet_config, lookup=True):
"""Create the CloudFormationConfiguration object."""
config = CloudFormationConfiguration('activities', bosslet_config)
names = bosslet_config.names
keypair = bosslet_config.SSH_KEY
session = bosslet_config.session
vpc_id = config.find_vpc()
sgs = aws.sg_lookup_all(session, vpc_id)
internal_subnets, _ = config.find_all_subnets()
internal_subnets_asg, _ = config.find_all_subnets(compatibility='asg')
topic_arn = aws.sns_topic_lookup(session, bosslet_config.ALERT_TOPIC)
if topic_arn is None:
raise MissingResourceError('SNS topic', bosslet_config.ALERT_TOPIC)
event_data = {
"lambda-name": "delete_lambda",
"db": names.endpoint_db.rds,
"meta-db": names.meta.ddb,
"s3-index-table": names.s3_index.ddb,
"id-index-table": names.id_index.ddb,
"id-count-table": names.id_count_index.ddb,
"cuboid_bucket": names.cuboid_bucket.s3,
"delete_bucket": names.delete_bucket.s3,
"topic-arn": topic_arn,
"query-deletes-sfn-name": names.query_deletes.sfn,
"delete-sfn-name": names.delete_cuboid.sfn,
"delete-exp-sfn-name": names.delete_experiment.sfn,
"delete-coord-frame-sfn-name": names.delete_coord_frame.sfn,
"delete-coll-sfn-name": names.delete_collection.sfn,
}
role_arn = aws.role_arn_lookup(session, "events_for_delete_lambda")
multi_lambda = names.multi_lambda.lambda_
if lookup:
lambda_arn = aws.lambda_arn_lookup(session, multi_lambda)
else:
lambda_arn = None
target_list = [{
"Arn": lambda_arn,
"Id": multi_lambda,
"Input": json.dumps(event_data)
}]
schedule_expression = "cron(1 6-11/1 ? * TUE-FRI *)"
#schedule_expression = "cron(0/2 * * * ? *)" # testing fire every two minutes
config.add_event_rule("DeleteEventRule",
# XXX What type for event rules?
names.delete_event_rule.dns,
role_arn=role_arn,
schedule_expression=schedule_expression,
target_list=target_list,
state='DISABLED') # Disabled until new delete is finished.
# Events have to be given permission to run lambda.
config.add_lambda_permission('DeleteRulePerm',
multi_lambda,
principal='events.amazonaws.com',
source=Arn('DeleteEventRule'))
user_data = UserData()
user_data["system"]["fqdn"] = names.activities.dns
user_data["system"]["type"] = "activities"
user_data["aws"]["db"] = names.endpoint_db.rds
user_data["aws"]["cache"] = names.cache.redis
user_data["aws"]["cache-state"] = names.cache_state.redis
user_data["aws"]["cache-db"] = "0"
user_data["aws"]["cache-state-db"] = "0"
user_data["aws"]["meta-db"] = names.meta.ddb
user_data["aws"]["cuboid_bucket"] = names.cuboid_bucket.s3
user_data["aws"]["tile_bucket"] = names.tile_bucket.s3
user_data["aws"]["ingest_bucket"] = names.ingest_bucket.s3
user_data["aws"]["s3-index-table"] = names.s3_index.ddb
user_data["aws"]["tile-index-table"] = names.tile_index.ddb
user_data["aws"]["id-index-table"] = names.id_index.ddb
user_data["aws"]["id-count-table"] = names.id_count_index.ddb
user_data["aws"]["max_task_id_suffix"] = str(const.MAX_TASK_ID_SUFFIX)
user_data["aws"]["tile_ingest_lambda"] = names.tile_ingest.lambda_
user_data["aws"]["tile_uploaded_lambda"] = names.tile_uploaded.lambda_
config.add_autoscale_group("Activities",
names.activities.dns,
aws.ami_lookup(bosslet_config, names.activities.ami),
keypair,
subnets=internal_subnets_asg,
type_=const.ACTIVITIES_TYPE,
security_groups=[sgs[names.internal.sg]],
user_data=str(user_data),
role=aws.instance_profile_arn_lookup(session, "activities"),
min=1,
max=1)
config.add_lambda("IngestLambda",
names.ingest_lambda.lambda_,
aws.role_arn_lookup(session, 'IngestQueueUpload'),
const.INGEST_LAMBDA,
handler="index.handler",
timeout=60 * 5,
runtime='python3.6',
memory=3008)
config.add_lambda_permission("IngestLambdaExecute", Ref("IngestLambda"))
# Downsample / Resolution Hierarchy support
lambda_role = aws.role_arn_lookup(session, "lambda_resolution_hierarchy")
config.add_lambda("DownsampleVolumeLambda",
names.downsample_volume.lambda_,
lambda_role,
handler="downsample_volume.handler",
timeout=120,
memory=1024,
dlq = Ref('DownsampleDLQ'))
start_sfn_lambda_role = aws.role_arn_lookup(session, 'StartStepFcnLambdaRole')
config.add_lambda("startSfnLambda",
names.start_sfn.lambda_,
start_sfn_lambda_role,
handler="start_sfn_lambda.handler",
timeout=60,
memory=128)
# This dead letter queue behavior uses a lambda to put failed lambda
# executions into a dlqs created specifically for each downsample job.
# There is a separate dlq for each resolution.
config.add_sns_topic("DownsampleDLQ",
names.downsample_dlq.sns,
names.downsample_dlq.sns,
[('lambda', Arn('DownsampleDLQLambda'))])
config.add_lambda('DownsampleDLQLambda',
names.downsample_dlq.lambda_,
lambda_role,
const.DOWNSAMPLE_DLQ_LAMBDA,
handler='index.handler',
runtime='python3.7',
timeout=10)
config.add_lambda_permission('DownsampleDLQLambdaExecute',
Ref('DownsampleDLQLambda'))
return config
def create_config(session, domain):
"""Create the CloudFormationConfiguration object.
:arg session used to perform lookups
:arg domain DNS name of vpc
"""
config = CloudFormationConfiguration('cloudwatch', domain)
names = AWSNames(domain)
vpc_id = config.find_vpc(session)
lambda_subnets, _ = config.find_all_availability_zones(session, lambda_compatible_only=True)
internal_sg = aws.sg_lookup(session, vpc_id, names.internal)
loadbalancer_name = names.endpoint_elb
if not aws.lb_lookup(session, loadbalancer_name):
raise Exception("Invalid load balancer name: " + loadbalancer_name)
# TODO Test that MailingListTopic is working.
production_mailing_list = const.PRODUCTION_MAILING_LIST
mailing_list_arn = aws.sns_topic_lookup(session, production_mailing_list)
if mailing_list_arn is None:
#config.add_sns_topic("topicList", production_mailing_list)
msg = "MailingList {} needs to be created before running config"
raise Exception(msg.format(const.PRODUCTION_MAILING_LIST))
config.add_cloudwatch(loadbalancer_name, [mailing_list_arn])
lambda_role = aws.role_arn_lookup(session, 'VaultConsulHealthChecker')
config.add_arg(Arg.String(
'VaultConsulHealthChecker', lambda_role,
'IAM role for vault/consul health check.' + domain))
config.add_lambda('VaultLambda',
names.vault_monitor,
description='Check health of vault instances.',
timeout=30,
role=Ref('VaultConsulHealthChecker'),
security_groups=[internal_sg],
subnets=lambda_subnets,
handler='index.lambda_handler',
file=const.VAULT_LAMBDA)
config.add_lambda('ConsulLambda',
names.consul_monitor,
description='Check health of vault instances.',
timeout=30,
role=Ref('VaultConsulHealthChecker'),
security_groups=[internal_sg],
subnets=lambda_subnets,
handler='index.lambda_handler',
file=const.CONSUL_LAMBDA)
# Lambda input data
json_str = json.dumps({
'vpc_id': vpc_id,
'vpc_name': domain,
'topic_arn': mailing_list_arn,
})
config.add_cloudwatch_rule('VaultConsulCheck',
name=names.vault_consul_check,
description='Check health of vault and consul instances.',
targets=[
{
'Arn': Arn('VaultLambda'),
'Id': names.vault_monitor,
'Input': json_str
},
{
'Arn': Arn('ConsulLambda'),
'Id': names.consul_monitor,
'Input': json_str
},
],
schedule='rate(1 minute)',
depends_on=['VaultLambda', 'ConsulLambda'])
config.add_lambda_permission('VaultPerms',
names.vault_monitor,
principal='events.amazonaws.com',
source=Arn('VaultConsulCheck'))
config.add_lambda_permission('ConsulPerms',
names.consul_monitor,
principal='events.amazonaws.com',
source=Arn('VaultConsulCheck'))
return config
def create_config(session, domain):
"""Create the CloudFormationConfiguration object."""
config = CloudFormationConfiguration('activities', domain, const.REGION)
names = AWSNames(domain)
global keypair
keypair = aws.keypair_lookup(session)
vpc_id = config.find_vpc(session)
sgs = aws.sg_lookup_all(session, vpc_id)
internal_subnets, _ = config.find_all_availability_zones(session)
internal_subnets_lambda, _ = config.find_all_availability_zones(
session, lambda_compatible_only=True)
topic_arn = aws.sns_topic_lookup(session, "ProductionMicronsMailingList")
event_data = {
"lambda-name": "delete_lambda",
"db": names.endpoint_db,
"meta-db": names.meta,
"s3-index-table": names.s3_index,
"id-index-table": names.id_index,
"id-count-table": names.id_count_index,
"cuboid_bucket": names.cuboid_bucket,
"delete_bucket": names.delete_bucket,
"topic-arn": topic_arn,
"query-deletes-sfn-name": names.query_deletes,
"delete-sfn-name": names.delete_cuboid,
"delete-exp-sfn-name": names.delete_experiment,
"delete-coord-frame-sfn-name": names.delete_coord_frame,
"delete-coll-sfn-name": names.delete_collection
}
role_arn = aws.role_arn_lookup(session, "events_for_delete_lambda")
multi_lambda = names.multi_lambda
lambda_arn = aws.lambda_arn_lookup(session, multi_lambda)
target_list = [{
"Arn": lambda_arn,
"Id": multi_lambda,
"Input": json.dumps(event_data)
}]
schedule_expression = "cron(1 6-11/1 ? * TUE-FRI *)"
#schedule_expression = "cron(0/2 * * * ? *)" # testing fire every two minutes
config.add_event_rule("DeleteEventRule",
names.delete_event_rule,
role_arn=role_arn,
schedule_expression=schedule_expression,
target_list=target_list,
description=None)
# Events have to be given permission to run lambda.
config.add_lambda_permission('DeleteRulePerm',
multi_lambda,
principal='events.amazonaws.com',
source=Arn('DeleteEventRule'))
user_data = UserData()
user_data["system"]["fqdn"] = names.activities
user_data["system"]["type"] = "activities"
user_data["aws"]["db"] = names.endpoint_db
user_data["aws"]["cache"] = names.cache
user_data["aws"]["cache-state"] = names.cache_state
user_data["aws"]["cache-db"] = "0"
user_data["aws"]["cache-state-db"] = "0"
user_data["aws"]["meta-db"] = names.meta
user_data["aws"]["cuboid_bucket"] = names.cuboid_bucket
user_data["aws"]["tile_bucket"] = names.tile_bucket
user_data["aws"]["ingest_bucket"] = names.ingest_bucket
user_data["aws"]["s3-index-table"] = names.s3_index
user_data["aws"]["tile-index-table"] = names.tile_index
user_data["aws"]["id-index-table"] = names.id_index
user_data["aws"]["id-count-table"] = names.id_count_index
config.add_autoscale_group("Activities",
names.activities,
aws.ami_lookup(session, 'activities.boss'),
keypair,
subnets=internal_subnets_lambda,
type_=const.ACTIVITIES_TYPE,
security_groups=[sgs[names.internal]],
user_data=str(user_data),
role=aws.instance_profile_arn_lookup(
session, "activities"),
min=1,
max=1)
config.add_lambda("IngestLambda",
names.ingest_lambda,
aws.role_arn_lookup(session, 'IngestQueueUpload'),
const.INGEST_LAMBDA,
handler="index.handler",
timeout=60 * 5)
config.add_lambda_permission("IngestLambdaExecute", Ref("IngestLambda"))
return config
def delete(session, domain):
names = AWSNames(domain)
aws.route53_delete_records(session, domain, names.endpoint)
aws.sqs_delete_all(session, domain)
aws.policy_delete_all(session, domain, '/ingest/')
CloudFormationConfiguration('api', domain).delete(session)
def create_config(session, domain, keypair=None, db_config={}):
"""
Create the CloudFormationConfiguration object.
Args:
session: amazon session object
domain (string): domain of the stack being created
keypair: keypair used to by instances being created
db_config (dict): information needed by rds
Returns: the config for the Cloud Formation stack
"""
names = AWSNames(domain)
# Lookup IAM Role and SNS Topic ARNs for used later in the config
endpoint_role_arn = aws.role_arn_lookup(session, "endpoint")
cachemanager_role_arn = aws.role_arn_lookup(session, 'cachemanager')
dns_arn = aws.sns_topic_lookup(session, names.dns.replace(".", "-"))
if dns_arn is None:
raise Exception("SNS topic named dns." + domain + " does not exist.")
mailing_list_arn = aws.sns_topic_lookup(session, const.PRODUCTION_MAILING_LIST)
if mailing_list_arn is None:
msg = "MailingList {} needs to be created before running config".format(const.PRODUCTION_MAILING_LIST)
raise Exception(msg)
# Configure Vault and create the user data config that the endpoint will
# use for connecting to Vault and the DB instance
user_data = UserData()
user_data["system"]["fqdn"] = names.endpoint
user_data["system"]["type"] = "endpoint"
user_data["aws"]["db"] = names.endpoint_db
user_data["aws"]["cache"] = names.cache
user_data["aws"]["cache-state"] = names.cache_state
## cache-db and cache-stat-db need to be in user_data for lambda to access them.
user_data["aws"]["cache-db"] = "0"
user_data["aws"]["cache-state-db"] = "0"
user_data["aws"]["meta-db"] = names.meta
# Use CloudFormation's Ref function so that queues' URLs are placed into
# the Boss config file.
user_data["aws"]["s3-flush-queue"] = str(Ref(names.s3flush_queue)) # str(Ref("S3FlushQueue")) DP XXX
user_data["aws"]["s3-flush-deadletter-queue"] = str(Ref(names.deadletter_queue)) #str(Ref("DeadLetterQueue")) DP XXX
user_data["aws"]["cuboid_bucket"] = names.cuboid_bucket
user_data["aws"]["tile_bucket"] = names.tile_bucket
user_data["aws"]["ingest_bucket"] = names.ingest_bucket
user_data["aws"]["s3-index-table"] = names.s3_index
user_data["aws"]["tile-index-table"] = names.tile_index
user_data["aws"]["id-index-table"] = names.id_index
user_data["aws"]["id-count-table"] = names.id_count_index
user_data["aws"]["prod_mailing_list"] = mailing_list_arn
user_data["auth"]["OIDC_VERIFY_SSL"] = 'True'
user_data["lambda"]["flush_function"] = names.multi_lambda
user_data["lambda"]["page_in_function"] = names.multi_lambda
user_data["lambda"]["ingest_function"] = names.multi_lambda
user_data['sfn']['populate_upload_queue'] = names.ingest_queue_populate
user_data['sfn']['upload_sfn'] = names.ingest_queue_upload
user_data['sfn']['downsample_sfn'] = names.resolution_hierarchy
# Prepare user data for parsing by CloudFormation.
parsed_user_data = { "Fn::Join" : ["", user_data.format_for_cloudformation()]}
config = CloudFormationConfiguration('api', domain, const.REGION)
vpc_id = config.find_vpc(session)
az_subnets, external_subnets = config.find_all_availability_zones(session)
az_subnets_lambda, external_subnets_lambda = config.find_all_availability_zones(session, lambda_compatible_only=True)
sgs = aws.sg_lookup_all(session, vpc_id)
# DP XXX: hack until we can get productio updated correctly
config.add_security_group('AllHTTPSSecurityGroup', 'https.' + domain, [('tcp', '443', '443', '0.0.0.0/0')])
sgs[names.https] = Ref('AllHTTPSSecurityGroup')
# Create SQS queues and apply access control policies.
#config.add_sqs_queue("DeadLetterQueue", names.deadletter_queue, 30, 20160) DP XXX
config.add_sqs_queue(names.deadletter_queue, names.deadletter_queue, 30, 20160)
max_receives = 3
#config.add_sqs_queue("S3FlushQueue", DP XXX
config.add_sqs_queue(names.s3flush_queue,
names.s3flush_queue,
30,
dead=(Arn(names.deadletter_queue), max_receives))
config.add_sqs_policy("sqsEndpointPolicy", 'sqsEndpointPolicy', # DP XXX
[Ref(names.deadletter_queue), Ref(names.s3flush_queue)],
endpoint_role_arn)
config.add_sqs_policy("sqsCachemgrPolicy", 'sqsCachemgrPolicy', # DP XXX
[Ref(names.deadletter_queue), Ref(names.s3flush_queue)],
cachemanager_role_arn)
# Create the endpoint ASG, ELB, and RDS instance
config.add_autoscale_group("Endpoint",
names.endpoint,
aws.ami_lookup(session, "endpoint.boss"),
keypair,
subnets=az_subnets_lambda,
type_=const.ENDPOINT_TYPE,
security_groups=[sgs[names.internal]],
user_data=parsed_user_data,
min=const.ENDPOINT_CLUSTER_MIN,
max=const.ENDPOINT_CLUSTER_MAX,
elb=Ref("EndpointLoadBalancer"),
notifications=dns_arn,
role=aws.instance_profile_arn_lookup(session, 'endpoint'),
health_check_grace_period=90,
detailed_monitoring=True,
depends_on=["EndpointLoadBalancer", "EndpointDB"])
cert = aws.cert_arn_lookup(session, names.public_dns("api"))
config.add_loadbalancer("EndpointLoadBalancer",
names.endpoint_elb,
[("443", "80", "HTTPS", cert)],
subnets=external_subnets_lambda,
security_groups=[sgs[names.internal], sgs[names.https]],
public=True)
# Endpoint servers are not CPU bound typically, so react quickly to load
config.add_autoscale_policy("EndpointScaleUp",
Ref("Endpoint"),
adjustments=[
(0.0, 10, 1), # 12% - 22% Utilization add 1 instance
(10, None, 2) # Above 22% Utilization add 2 instances
],
alarms=[
("CPUUtilization", "Maximum", "GreaterThanThreshold", "12")
],
period=1)
config.add_autoscale_policy("EndpointScaleDown",
Ref("Endpoint"),
adjustments=[
(None, 0.0, -1), # Under 1.5% Utilization remove 1 instance
],
alarms=[
("CPUUtilization", "Average", "LessThanThreshold", "1.5")
],
period=50)
config.add_rds_db("EndpointDB",
names.endpoint_db,
db_config.get("port"),
db_config.get("name"),
db_config.get("user"),
db_config.get("password"),
az_subnets,
type_ = const.RDS_TYPE,
security_groups=[sgs[names.internal]])
# Create the Meta, s3Index, tileIndex, annotation Dynamo tables
with open(const.DYNAMO_METADATA_SCHEMA, 'r') as fh:
dynamo_cfg = json.load(fh)
config.add_dynamo_table_from_json("EndpointMetaDB", names.meta, **dynamo_cfg)
with open(const.DYNAMO_S3_INDEX_SCHEMA, 'r') as s3fh:
dynamo_s3_cfg = json.load(s3fh)
config.add_dynamo_table_from_json('s3Index', names.s3_index, **dynamo_s3_cfg) # DP XXX
with open(const.DYNAMO_TILE_INDEX_SCHEMA, 'r') as tilefh:
dynamo_tile_cfg = json.load(tilefh)
config.add_dynamo_table_from_json('tileIndex', names.tile_index, **dynamo_tile_cfg) # DP XXX
with open(const.DYNAMO_ID_INDEX_SCHEMA, 'r') as id_ind_fh:
dynamo_id_ind__cfg = json.load(id_ind_fh)
config.add_dynamo_table_from_json('idIndIndex', names.id_index, **dynamo_id_ind__cfg) # DP XXX
with open(const.DYNAMO_ID_COUNT_SCHEMA, 'r') as id_count_fh:
dynamo_id_count_cfg = json.load(id_count_fh)
config.add_dynamo_table_from_json('idCountIndex', names.id_count_index, **dynamo_id_count_cfg) # DP XXX
return config
def delete(bosslet_config):
CloudFormationConfiguration('activities', bosslet_config).delete()
delete_sfns(bosslet_config)
def delete(bosslet_config):
CloudFormationConfiguration('idindexing', bosslet_config).delete()
delete_sfns(bosslet_config)
def create_config(bosslet_config, user_data=None):
"""
Create the CloudFormationConfiguration object.
Args:
bosslet_config (BossConfiguration): target bosslet
user_data (UserData): information used by the endpoint instance and vault. Data will be run through the CloudFormation Fn::Join template intrinsic function so other template intrinsic functions used in the user_data will be parsed and executed.
Returns: the config for the Cloud Formation stack
"""
# Prepare user data for parsing by CloudFormation.
if user_data is not None:
parsed_user_data = { "Fn::Join" : ["", user_data.format_for_cloudformation()]}
else:
parsed_user_data = user_data
keypair = bosslet_config.SSH_KEY
session = bosslet_config.session
names = bosslet_config.names
config = CloudFormationConfiguration("cachedb", bosslet_config)
vpc_id = config.find_vpc()
#####
# TODO: When CF config files are refactored for multi-account support
# the creation of _all_ subnets should be moved into core.
# AWS doesn't charge for the VPC or subnets, so it doesn't
# increase cost and cleans up subnet creation
# Create several subnets for all the lambdas to use.
internal_route_table_id = aws.rt_lookup(session, vpc_id, names.internal.rt)
lambda_subnets = config.add_all_lambda_subnets()
for lambda_subnet in lambda_subnets:
key = lambda_subnet['Ref']
config.add_route_table_association(key + "RTA",
internal_route_table_id,
lambda_subnet)
# Create a custom resource to help delete ENIs from lambdas
# DP NOTE: After deleting a lambda the ENIs may stick around for while, causing the stack delete to fail
# See https://stackoverflow.com/a/41310289
config.add_arg(Arg.String('StackName', config.stack_name))
config.add_arg(Arg.String('DeleteENILambda', aws.lambda_arn_lookup(session, names.delete_eni.lambda_)))
config.add_custom_resource('DeleteENI', 'DeleteENI', Ref('DeleteENILambda'), StackName = Ref('StackName'))
# Lookup the External Subnet, Internal Security Group IDs that are
# needed by other resources
internal_subnet_id = aws.subnet_id_lookup(session, names.internal.subnet)
config.add_arg(Arg.Subnet("InternalSubnet",
internal_subnet_id,
"ID of Internal Subnet to create resources in"))
internal_sg_id = aws.sg_lookup(session, vpc_id, names.internal.sg)
config.add_arg(Arg.SecurityGroup("InternalSecurityGroup",
internal_sg_id,
"ID of internal Security Group"))
role = aws.role_arn_lookup(session, "lambda_cache_execution")
config.add_arg(Arg.String("LambdaCacheExecutionRole", role,
"IAM role for " + names.multi_lambda.lambda_))
cuboid_import_role = aws.role_arn_lookup(session, CUBOID_IMPORT_ROLE)
config.add_arg(Arg.String(CUBOID_IMPORT_ROLE, cuboid_import_role,
"IAM role for cuboidImport"))
config.add_capabilities(['CAPABILITY_IAM'])
cuboid_bucket_name = names.cuboid_bucket.s3
if not aws.s3_bucket_exists(session, cuboid_bucket_name):
config.add_s3_bucket("cuboidBucket", cuboid_bucket_name)
config.add_s3_bucket_policy(
"cuboidBucketPolicy", cuboid_bucket_name,
['s3:GetObject', 's3:PutObject'],
{ 'AWS': role})
config.append_s3_bucket_policy(
"cuboidBucketPolicy", cuboid_bucket_name,
['s3:PutObject'], { 'AWS': cuboid_import_role})
delete_bucket_name = names.delete_bucket.s3
if not aws.s3_bucket_exists(session, delete_bucket_name):
config.add_s3_bucket("deleteBucket", delete_bucket_name)
config.add_s3_bucket_policy(
"deleteBucketPolicy", delete_bucket_name,
['s3:GetObject', 's3:PutObject'],
{ 'AWS': role})
tile_bucket_name = names.tile_bucket.s3
if not aws.s3_bucket_exists(session, tile_bucket_name):
life_cycle_cfg = get_cf_bucket_life_cycle_rules()
config.add_s3_bucket(
"tileBucket", tile_bucket_name, life_cycle_config=life_cycle_cfg)
config.add_s3_bucket_policy(
"tileBucketPolicy", tile_bucket_name,
['s3:GetObject', 's3:PutObject'],
{ 'AWS': role})
# The ingest bucket is a staging area for cuboids uploaded during volumetric ingest.
creating_ingest_bucket = False
ingest_bucket_name = names.ingest_bucket.s3
if not aws.s3_bucket_exists(session, ingest_bucket_name):
creating_ingest_bucket = True
ing_bucket_life_cycle_cfg = get_cf_bucket_life_cycle_rules()
config.add_s3_bucket("ingestBucket", ingest_bucket_name,
life_cycle_config=ing_bucket_life_cycle_cfg)
config.add_s3_bucket_policy(
"ingestBucketPolicy", ingest_bucket_name,
['s3:GetObject', 's3:PutObject', 's3:PutObjectTagging'],
{ 'AWS': cuboid_import_role})
config.add_ec2_instance("CacheManager",
names.cachemanager.dns,
aws.ami_lookup(bosslet_config, names.cachemanager.ami),
keypair,
subnet=Ref("InternalSubnet"),
public_ip=False,
type_=const.CACHE_MANAGER_TYPE,
security_groups=[Ref("InternalSecurityGroup")],
user_data=parsed_user_data,
role="cachemanager")
config.add_sqs_queue(
names.ingest_cleanup_dlq.sqs, names.ingest_cleanup_dlq.sqs, 30, 20160)
config.add_sqs_queue(
names.cuboid_import_dlq.sqs, names.cuboid_import_dlq.sqs, 30, 20160)
config.add_sqs_policy('cuboidImportDlqPolicy', 'cuboidImportDlqPolicy',
[Ref(names.cuboid_import_dlq.sqs)], cuboid_import_role)
config.add_lambda("MultiLambda",
names.multi_lambda.lambda_,
Ref("LambdaCacheExecutionRole"),
handler='lambda_loader.handler',
timeout=120,
memory=1536,
security_groups=[Ref('InternalSecurityGroup')],
subnets=lambda_subnets)
config.add_lambda("TileUploadedLambda",
names.tile_uploaded.lambda_,
Ref("LambdaCacheExecutionRole"),
handler='tile_uploaded_lambda.handler',
timeout=5,
memory=1024)
config.add_lambda("TileIngestLambda",
names.tile_ingest.lambda_,
Ref("LambdaCacheExecutionRole"),
handler='tile_ingest_lambda.handler',
timeout=30,
memory=1536)
config.add_lambda("DeleteTileObjsLambda",
names.delete_tile_objs.lambda_,
Ref("LambdaCacheExecutionRole"),
handler='delete_tile_objs_lambda.handler',
timeout=90,
memory=128,
dlq=Arn(names.ingest_cleanup_dlq.sqs))
config.add_lambda("DeleteTileEntryLambda",
names.delete_tile_index_entry.lambda_,
Ref("LambdaCacheExecutionRole"),
handler='delete_tile_index_entry_lambda.handler',
timeout=90,
memory=128,
dlq=Arn(names.ingest_cleanup_dlq.sqs))
config.add_lambda("CuboidImportLambda",
names.cuboid_import_lambda.lambda_,
Ref(CUBOID_IMPORT_ROLE),
handler='cuboid_import_lambda.handler',
timeout=90,
memory=128,
dlq=Arn(names.cuboid_import_dlq.sqs))
config.add_lambda("VolumetricIngestLambda",
names.volumetric_ingest_queue_upload_lambda.lambda_,
Ref("LambdaCacheExecutionRole"),
handler='ingest_queue_upload_volumetric_lambda.handler',
timeout=120,
memory=1024)
if creating_ingest_bucket:
config.add_lambda_permission(
'ingestBucketInvokeCuboidImportLambda', names.cuboid_import_lambda.lambda_,
principal='s3.amazonaws.com', source={
'Fn::Join': [':', ['arn', 'aws', 's3', '', '', ingest_bucket_name]]}, #DP TODO: move into constants
depends_on=['ingestBucket', 'CuboidImportLambda']
)
else:
# NOTE: this permission doesn't seem to apply properly when doing a
# CloudFormation update. During testing, I had to manually apply this
# permission before the bucket trigger could be applied in post_init().
# Doing a CloudFormation delete followed by a create did not have a
# problem.
config.add_lambda_permission(
'ingestBucketInvokeCuboidImportLambda', names.cuboid_import_lambda.lambda_,
principal='s3.amazonaws.com', source={
'Fn::Join': [':', ['arn', 'aws', 's3', '', '', ingest_bucket_name]]},
depends_on='CuboidImportLambda'
)
# Add topic to indicating that the object store has been write locked.
# Now using "production mailing list" instead of separate write lock topic.
#config.add_sns_topic('WriteLock',
# names.write_lock_topic,
# names.write_lock,
# []) # TODO: add subscribers
return config
def create_config(bosslet_config):
"""Create the CloudFormationConfiguration object."""
config = CloudFormationConfiguration('idindexing', bosslet_config)
session = bosslet_config.session
names = bosslet_config.names
#topic_arn = aws.sns_topic_lookup(session, "ProductionMicronsMailingList")
role = aws.role_arn_lookup(session, "lambda_cache_execution")
config.add_arg(
Arg.String("LambdaCacheExecutionRole", role,
"IAM role for " + names.multi_lambda.lambda_))
def add_lambda(key, name, handler, timeout, memory):
"""A method for defining the common arguments for adding a lambda"""
config.add_lambda(key,
name,
Ref('LambdaCacheExecutionRole'),
handler=handler,
timeout=timeout,
memory=memory)
add_lambda("indexS3WriterLambda",
names.index_s3_writer.lambda_,
"write_s3_index_lambda.handler",
timeout=120,
memory=1024)
add_lambda("indexFanoutIdWriterLambda",
names.index_fanout_id_writer.lambda_,
"fanout_write_id_index_lambda.handler",
timeout=120,
memory=256)
add_lambda("indexWriteIdLambda",
names.index_write_id.lambda_,
"write_id_index_lambda.handler",
timeout=120,
memory=512)
add_lambda("indexWriteFailedLambda",
names.index_write_failed.lambda_,
"write_index_failed_lambda.handler",
timeout=60,
memory=128)
add_lambda("indexFindCuboidsLambda",
names.index_find_cuboids.lambda_,
"index_find_cuboids_lambda.handler",
timeout=120,
memory=256)
add_lambda("indexFanoutEnqueueCuboidsKeysLambda",
names.index_fanout_enqueue_cuboid_keys.lambda_,
"fanout_enqueue_cuboid_keys_lambda.handler",
timeout=120,
memory=256)
add_lambda("indexBatchEnqueueCuboidsLambda",
names.index_batch_enqueue_cuboids.lambda_,
"batch_enqueue_cuboids_lambda.handler",
timeout=60,
memory=128)
add_lambda("indexFanoutDequeueCuboidKeysLambda",
names.index_fanout_dequeue_cuboid_keys.lambda_,
"fanout_dequeue_cuboid_keys_lambda.handler",
timeout=60,
memory=128)
add_lambda("indexDequeueCuboidKeysLambda",
names.index_dequeue_cuboid_keys.lambda_,
"dequeue_cuboid_keys_lambda.handler",
timeout=60,
memory=128)
add_lambda("indexGetNumCuboidKeysMsgsLambda",
names.index_get_num_cuboid_keys_msgs.lambda_,
"get_num_msgs_cuboid_keys_queue_lambda.handler",
timeout=60,
memory=128)
add_lambda("indexCheckForThrottlingLambda",
names.index_check_for_throttling.lambda_,
"check_for_index_throttling_lambda.handler",
timeout=60,
memory=128)
add_lambda("indexInvokeIndexSupervisorLambda",
names.index_invoke_index_supervisor.lambda_,
"invoke_index_supervisor_lambda.handler",
timeout=60,
memory=128)
add_lambda("indexSplitCuboidsLambda",
names.index_split_cuboids.lambda_,
"split_cuboids_lambda.handler",
timeout=120,
memory=128)
add_lambda("indexLoadIdsFromS3Lambda",
names.index_load_ids_from_s3.lambda_,
"load_ids_from_s3_lambda.handler",
timeout=120,
memory=128)
return config
def create_config(session, domain):
"""Create the CloudFormationConfiguration object."""
config = CloudFormationConfiguration('core', domain, const.REGION)
names = AWSNames(domain)
global keypair
keypair = aws.keypair_lookup(session)
config.add_vpc()
# Create the internal and external subnets
config.add_subnet('InternalSubnet', names.subnet('internal'))
config.add_subnet('ExternalSubnet', names.subnet('external'))
internal_subnets, external_subnets = config.add_all_azs(session)
# it seems that both Lambdas and ASGs needs lambda_compatible_only subnets.
internal_subnets_lambda, external_subnets_lambda = config.add_all_azs(session, lambda_compatible_only=True)
config.add_ec2_instance("Bastion",
names.bastion,
aws.ami_lookup(session, const.BASTION_AMI),
keypair,
subnet = Ref("ExternalSubnet"),
public_ip = True,
user_data = const.BASTION_USER_DATA,
security_groups = [Ref("InternalSecurityGroup"), Ref("BastionSecurityGroup")],
depends_on = "AttachInternetGateway")
user_data = UserData()
user_data["system"]["fqdn"] = names.consul
user_data["system"]["type"] = "consul"
user_data["consul"]["cluster"] = str(get_scenario(const.CONSUL_CLUSTER_SIZE))
config.add_autoscale_group("Consul",
names.consul,
aws.ami_lookup(session, "consul.boss"),
keypair,
subnets = internal_subnets_lambda,
security_groups = [Ref("InternalSecurityGroup")],
user_data = str(user_data),
min = const.CONSUL_CLUSTER_SIZE,
max = const.CONSUL_CLUSTER_SIZE,
notifications = Ref("DNSSNS"),
role = aws.instance_profile_arn_lookup(session, 'consul'),
support_update = False, # Update will restart the instances manually
depends_on = ["DNSLambda", "DNSSNS", "DNSLambdaExecute"])
user_data = UserData()
user_data["system"]["fqdn"] = names.vault
user_data["system"]["type"] = "vault"
config.add_autoscale_group("Vault",
names.vault,
aws.ami_lookup(session, "vault.boss"),
keypair,
subnets = internal_subnets_lambda,
security_groups = [Ref("InternalSecurityGroup")],
user_data = str(user_data),
min = const.VAULT_CLUSTER_SIZE,
max = const.VAULT_CLUSTER_SIZE,
notifications = Ref("DNSSNS"),
depends_on = ["Consul", "DNSLambda", "DNSSNS", "DNSLambdaExecute"])
user_data = UserData()
user_data["system"]["fqdn"] = names.auth
user_data["system"]["type"] = "auth"
deps = ["AuthSecurityGroup",
"AttachInternetGateway",
"DNSLambda",
"DNSSNS",
"DNSLambdaExecute"]
SCENARIO = os.environ["SCENARIO"]
USE_DB = SCENARIO in ("production", "ha-development",)
# Problem: If development scenario uses a local DB. If the auth server crashes
# and is auto restarted by the autoscale group then the new auth server
# will not have any of the previous configuration, because the old DB
# was lost. Using an RDS for development fixes this at the cost of having
# the core config taking longer to launch.
if USE_DB:
deps.append("AuthDB")
user_data["aws"]["db"] = "keycloak" # flag for init script for which config to use
cert = aws.cert_arn_lookup(session, names.public_dns('auth'))
create_asg_elb(config,
"Auth",
names.auth,
aws.ami_lookup(session, "auth.boss"),
keypair,
str(user_data),
const.AUTH_CLUSTER_SIZE,
internal_subnets_lambda,
external_subnets_lambda,
[("443", "8080", "HTTPS", cert)],
"HTTP:8080/index.html",
sgs = [Ref("AuthSecurityGroup")],
type_=const.AUTH_TYPE,
depends_on=deps)
if USE_DB:
config.add_rds_db("AuthDB",
names.auth_db,
"3306",
"keycloak",
"keycloak",
"keycloak",
internal_subnets,
type_ = "db.t2.micro",
security_groups = [Ref("InternalSecurityGroup")])
config.add_lambda("DNSLambda",
names.dns,
aws.role_arn_lookup(session, 'UpdateRoute53'),
const.DNS_LAMBDA,
handler="index.handler",
timeout=10,
depends_on="DNSZone")
config.add_lambda_permission("DNSLambdaExecute", Ref("DNSLambda"))
config.add_sns_topic("DNSSNS",
names.dns,
names.dns,
[("lambda", Arn("DNSLambda"))])
config.add_security_group("InternalSecurityGroup",
names.internal,
[("-1", "-1", "-1", "10.0.0.0/8")])
# Allow SSH access to bastion from anywhere
config.add_security_group("BastionSecurityGroup",
names.ssh,
[("tcp", "22", "22", const.INCOMING_SUBNET)])
config.add_security_group("AuthSecurityGroup",
#names.https, DP XXX: hack until we can get production updated correctly
names.auth,
[("tcp", "443", "443", "0.0.0.0/0")])
# Create the internal route table to route traffic to the NAT Bastion
all_internal_subnets = internal_subnets.copy()
all_internal_subnets.append(Ref("InternalSubnet"))
config.add_route_table("InternalRouteTable",
names.internal,
subnets = all_internal_subnets)
config.add_route_table_route("InternalNatRoute",
Ref("InternalRouteTable"),
nat = Ref("NAT"),
depends_on = "NAT")
# Create the internet gateway and internet router
all_external_subnets = external_subnets.copy()
all_external_subnets.append(Ref("ExternalSubnet"))
config.add_route_table("InternetRouteTable",
names.internet,
subnets = all_external_subnets)
config.add_route_table_route("InternetRoute",
Ref("InternetRouteTable"),
gateway = Ref("InternetGateway"),
depends_on = "AttachInternetGateway")
config.add_internet_gateway("InternetGateway", names.internet)
config.add_endpoint("S3Endpoint", "s3", [Ref("InternalRouteTable"), Ref('InternetRouteTable')])
config.add_endpoint("DynamoDBEndpoint", "dynamodb", [Ref("InternalRouteTable"), Ref('InternetRouteTable')])
config.add_nat("NAT", Ref("ExternalSubnet"), depends_on="AttachInternetGateway")
return config
def create_config(bosslet_config):
"""Create the CloudFormationConfiguration object.
:arg session used to perform lookups
:arg domain DNS name of vpc
"""
config = CloudFormationConfiguration('cloudwatch', bosslet_config)
names = bosslet_config.names
session = bosslet_config.session
domain = bosslet_config.INTERNAL_DOMAIN
vpc_id = config.find_vpc()
lambda_subnets, _ = config.find_all_subnets(compatibility='lambda')
internal_sg = aws.sg_lookup(session, vpc_id, names.internal.sg)
loadbalancer_name = names.endpoint_elb.dns
if not aws.lb_lookup(session, loadbalancer_name):
raise MissingResourceError('ELB', loadbalancer_name)
# TODO Test that MailingListTopic is working.
production_mailing_list = bosslet_config.ALERT_TOPIC
mailing_list_arn = aws.sns_topic_lookup(session, production_mailing_list)
if mailing_list_arn is None:
raise MissingResourceError('SNS topic', bosslet_config.ALERT_TOPIC)
config.add_cloudwatch(loadbalancer_name, [mailing_list_arn])
lambda_role = aws.role_arn_lookup(session, 'VaultConsulHealthChecker')
config.add_arg(
Arg.String('VaultConsulHealthChecker', lambda_role,
'IAM role for vault/consul health check'))
config.add_lambda('VaultLambda',
names.vault_monitor.lambda_,
description='Check health of vault instances.',
timeout=30,
role=Ref('VaultConsulHealthChecker'),
security_groups=[internal_sg],
subnets=lambda_subnets,
handler='index.lambda_handler',
file=const.VAULT_LAMBDA)
# Lambda input data
json_str = json.dumps({
'hostname': names.vault.dns,
})
config.add_cloudwatch_rule('VaultCheck',
name=names.vault_check.cw,
description='Check health of vault instances.',
targets=[
{
'Arn': Arn('VaultLambda'),
'Id': names.vault_monitor.lambda_,
'Input': json_str
},
],
schedule='rate(2 minutes)',
depends_on=['VaultLambda'])
config.add_lambda_permission('VaultPerms',
names.vault_monitor.lambda_,
principal='events.amazonaws.com',
source=Arn('VaultCheck'))
return config
def create_config(session, domain):
"""Create the CloudFormationConfiguration object."""
config = CloudFormationConfiguration('activities', domain, const.REGION)
names = AWSNames(domain)
global keypair
keypair = aws.keypair_lookup(session)
vpc_id = config.find_vpc(session)
sgs = aws.sg_lookup_all(session, vpc_id)
internal_subnets, _ = config.find_all_availability_zones(session)
internal_subnets_lambda, _ = config.find_all_availability_zones(session, lambda_compatible_only=True)
topic_arn = aws.sns_topic_lookup(session, "ProductionMicronsMailingList")
event_data = {
"lambda-name": "delete_lambda",
"db": names.endpoint_db,
"meta-db": names.meta,
"s3-index-table": names.s3_index,
"id-index-table": names.id_index,
"id-count-table": names.id_count_index,
"cuboid_bucket": names.cuboid_bucket,
"delete_bucket": names.delete_bucket,
"topic-arn": topic_arn,
"query-deletes-sfn-name": names.query_deletes,
"delete-sfn-name": names.delete_cuboid,
"delete-exp-sfn-name": names.delete_experiment,
"delete-coord-frame-sfn-name": names.delete_coord_frame,
"delete-coll-sfn-name": names.delete_collection
}
role_arn = aws.role_arn_lookup(session, "events_for_delete_lambda")
multi_lambda = names.multi_lambda
lambda_arn = aws.lambda_arn_lookup(session, multi_lambda)
target_list = [{
"Arn": lambda_arn,
"Id": multi_lambda,
"Input": json.dumps(event_data)
}]
schedule_expression = "cron(1 6-11/1 ? * TUE-FRI *)"
#schedule_expression = "cron(0/2 * * * ? *)" # testing fire every two minutes
config.add_event_rule("DeleteEventRule", names.delete_event_rule, role_arn=role_arn,
schedule_expression=schedule_expression, target_list=target_list, description=None)
# Events have to be given permission to run lambda.
config.add_lambda_permission('DeleteRulePerm', multi_lambda, principal='events.amazonaws.com',
source=Arn('DeleteEventRule'))
user_data = UserData()
user_data["system"]["fqdn"] = names.activities
user_data["system"]["type"] = "activities"
user_data["aws"]["db"] = names.endpoint_db
user_data["aws"]["cache"] = names.cache
user_data["aws"]["cache-state"] = names.cache_state
user_data["aws"]["cache-db"] = "0"
user_data["aws"]["cache-state-db"] = "0"
user_data["aws"]["meta-db"] = names.meta
user_data["aws"]["cuboid_bucket"] = names.cuboid_bucket
user_data["aws"]["tile_bucket"] = names.tile_bucket
user_data["aws"]["ingest_bucket"] = names.ingest_bucket
user_data["aws"]["s3-index-table"] = names.s3_index
user_data["aws"]["tile-index-table"] = names.tile_index
user_data["aws"]["id-index-table"] = names.id_index
user_data["aws"]["id-count-table"] = names.id_count_index
user_data["aws"]["max_task_id_suffix"] = str(const.MAX_TASK_ID_SUFFIX)
config.add_autoscale_group("Activities",
names.activities,
aws.ami_lookup(session, 'activities.boss'),
keypair,
subnets=internal_subnets_lambda,
type_=const.ACTIVITIES_TYPE,
security_groups=[sgs[names.internal]],
user_data=str(user_data),
role=aws.instance_profile_arn_lookup(session, "activities"),
min=1,
max=1)
config.add_lambda("IngestLambda",
names.ingest_lambda,
aws.role_arn_lookup(session, 'IngestQueueUpload'),
const.INGEST_LAMBDA,
handler="index.handler",
timeout=60 * 5,
memory=3008)
config.add_lambda_permission("IngestLambdaExecute", Ref("IngestLambda"))
# Downsample / Resolution Hierarchy support
lambda_role = aws.role_arn_lookup(session, "lambda_resolution_hierarchy")
config.add_lambda("DownsampleVolumeLambda",
names.downsample_volume_lambda,
lambda_role,
s3=(aws.get_lambda_s3_bucket(session),
"multilambda.{}.zip".format(domain),
"downsample_volume.handler"),
timeout=120,
memory=1024,
runtime='python3.6',
dlq = Ref('DownsampleDLQ'))
config.add_sns_topic("DownsampleDLQ",
names.downsample_dlq,
names.downsample_dlq,
[('lambda', Arn('DownsampleDLQLambda'))])
config.add_lambda('DownsampleDLQLambda',
names.downsample_dlq,
lambda_role,
const.DOWNSAMPLE_DLQ_LAMBDA,
handler='index.handler',
timeout=10)
config.add_lambda_permission('DownsampleDLQLambdaExecute',
Ref('DownsampleDLQLambda'))
return config
def delete(session, domain):
# NOTE: CloudWatch logs for the DNS Lambda are not deleted
names = AWSNames(domain)
aws.route53_delete_records(session, domain, names.cache_manager)
CloudFormationConfiguration("cachedb", domain).delete(session)
def create_config(bosslet_config):
"""Create the CloudFormationConfiguration object."""
config = CloudFormationConfiguration('backup', bosslet_config)
names = bosslet_config.names
# DP NOTE: During implementation there was/is an availability zone that
# could not run the T2.Micro instances used by this Data Pipeline
azs = aws.azs_lookup(bosslet_config, 'datapipeline')
az = random.choice(azs)[1] + '-'
internal_subnet = aws.subnet_id_lookup(
bosslet_config.session, az + bosslet_config.names.internal.subnet)
backup_image = aws.ami_lookup(bosslet_config, names.backup.ami)[0]
s3_backup = "s3://" + names.backup.s3 + "/#{format(@scheduledStartTime, 'YYYY-ww')}"
s3_logs = "s3://" + names.backup.s3 + "/logs"
# DP TODO: Create all BOSS S3 buckets as part of the account setup
# as the Cloud Formation delete doesn't delete the bucket,
# making this a conditional add
BUCKET_DEPENDENCY = None # Needed as the pipelines try to execute when launched
if not aws.s3_bucket_exists(bosslet_config.session, names.backup.s3):
life_cycle = {
'Rules': [{
'Id': 'Delete Data',
'Status': 'Enabled',
'ExpirationInDays': 180, # ~6 Months
}]
}
encryption = {
'ServerSideEncryptionConfiguration': [{
'ServerSideEncryptionByDefault': {
'SSEAlgorithm': 'AES256'
}
}]
}
config.add_s3_bucket("BackupBucket",
names.backup.s3,
life_cycle_config=life_cycle,
encryption=encryption)
BUCKET_DEPENDENCY = "BackupBucket"
# Vault Backup
cmd = "/usr/local/bin/python3 ~/vault.py backup {}".format(
bosslet_config.INTERNAL_DOMAIN)
pipeline = DataPipeline(log_uri=s3_logs, resource_role="backup")
pipeline.add_shell_command("VaultBackup",
cmd,
destination=Ref("VaultBucket"),
runs_on=Ref("VaultInstance"))
pipeline.add_s3_bucket("VaultBucket", s3_backup + "/vault")
pipeline.add_ec2_instance("VaultInstance",
subnet=internal_subnet,
image=backup_image)
config.add_data_pipeline("VaultBackupPipeline",
"vault-backup." + bosslet_config.INTERNAL_DOMAIN,
pipeline.objects,
depends_on=BUCKET_DEPENDENCY)
# DynamoDB Backup
tables = {
"BossMeta": names.meta.ddb,
"S3Index": names.s3_index.ddb,
"TileIndex": names.tile_index.ddb,
"IdIndex": names.id_index.ddb,
"IdCountIndex": names.id_count_index.ddb,
"VaultData": names.vault.ddb,
}
pipeline = DataPipeline(log_uri=s3_logs)
pipeline.add_emr_cluster("BackupCluster", region=bosslet_config.REGION)
for name in tables:
table = tables[name]
pipeline.add_s3_bucket(name + "Bucket", s3_backup + "/DDB/" + table)
pipeline.add_ddb_table(name, table)
pipeline.add_emr_copy(name + "Copy",
Ref(name),
Ref(name + "Bucket"),
runs_on=Ref("BackupCluster"),
region=bosslet_config.REGION)
config.add_data_pipeline("DDBPipeline",
"dynamo-backup." + bosslet_config.INTERNAL_DOMAIN,
pipeline.objects,
depends_on=BUCKET_DEPENDENCY)
# Endpoint RDS Backup
pipeline = rds_copy(names.endpoint_db.rds, internal_subnet, backup_image,
s3_logs, s3_backup)
config.add_data_pipeline("EndpointPipeline",
"endpoint-backup." +
bosslet_config.INTERNAL_DOMAIN,
pipeline.objects,
depends_on=BUCKET_DEPENDENCY)
# Auth RDS Backup
if bosslet_config.AUTH_RDS:
pipeline = rds_copy(names.auth_db.rds, internal_subnet, backup_image,
s3_logs, s3_backup)
config.add_data_pipeline("AuthPipeline",
"auth-backup." +
bosslet_config.INTERNAL_DOMAIN,
pipeline.objects,
depends_on=BUCKET_DEPENDENCY)
return config
def create_config(session, domain, keypair=None, user_data=None):
"""
Create the CloudFormationConfiguration object.
Args:
session: amazon session object
domain: domain of the stack being created
keypair: keypair used to by instances being created
user_data (UserData): information used by the endpoint instance and vault. Data will be run through the CloudFormation Fn::Join template intrinsic function so other template intrinsic functions used in the user_data will be parsed and executed.
Returns: the config for the Cloud Formation stack
"""
# Prepare user data for parsing by CloudFormation.
if user_data is not None:
parsed_user_data = {
"Fn::Join": ["", user_data.format_for_cloudformation()]
}
else:
parsed_user_data = user_data
names = AWSNames(domain)
config = CloudFormationConfiguration("cachedb", domain, const.REGION)
vpc_id = config.find_vpc(session)
# Create several subnets for all the lambdas to use.
lambda_azs = aws.azs_lookup(session, lambda_compatible_only=True)
internal_route_table_id = aws.rt_lookup(session, vpc_id, names.internal)
print("AZs for lambda: " + str(lambda_azs))
lambda_subnets = []
for i in range(const.LAMBDA_SUBNETS):
key = 'LambdaSubnet{}'.format(i)
lambda_subnets.append(Ref(key))
config.add_subnet(key,
names.subnet('lambda{}'.format(i)),
az=lambda_azs[i % len(lambda_azs)][0])
config.add_route_table_association(key + "RTA",
internal_route_table_id, Ref(key))
# Lookup the External Subnet, Internal Security Group IDs that are
# needed by other resources
internal_subnet_id = aws.subnet_id_lookup(session,
names.subnet("internal"))
config.add_arg(
Arg.Subnet("InternalSubnet", internal_subnet_id,
"ID of Internal Subnet to create resources in"))
internal_sg_id = aws.sg_lookup(session, vpc_id, names.internal)
config.add_arg(
Arg.SecurityGroup("InternalSecurityGroup", internal_sg_id,
"ID of internal Security Group"))
role = aws.role_arn_lookup(session, "lambda_cache_execution")
config.add_arg(
Arg.String("LambdaCacheExecutionRole", role,
"IAM role for multilambda." + domain))
index_bucket_name = names.cuboid_bucket
if not aws.s3_bucket_exists(session, index_bucket_name):
config.add_s3_bucket("cuboidBucket", index_bucket_name)
config.add_s3_bucket_policy("cuboidBucketPolicy", index_bucket_name,
['s3:GetObject', 's3:PutObject'],
{'AWS': role})
delete_bucket_name = names.delete_bucket
if not aws.s3_bucket_exists(session, delete_bucket_name):
config.add_s3_bucket("deleteBucket", delete_bucket_name)
config.add_s3_bucket_policy("deleteBucketPolicy", delete_bucket_name,
['s3:GetObject', 's3:PutObject'],
{'AWS': role})
creating_tile_bucket = False
tile_bucket_name = names.tile_bucket
if not aws.s3_bucket_exists(session, tile_bucket_name):
creating_tile_bucket = True
config.add_s3_bucket("tileBucket", tile_bucket_name)
config.add_s3_bucket_policy("tileBucketPolicy", tile_bucket_name,
['s3:GetObject', 's3:PutObject'],
{'AWS': role})
ingest_bucket_name = names.ingest_bucket
if not aws.s3_bucket_exists(session, ingest_bucket_name):
config.add_s3_bucket("ingestBucket", ingest_bucket_name)
config.add_s3_bucket_policy("ingestBucketPolicy", ingest_bucket_name,
['s3:GetObject', 's3:PutObject'],
{'AWS': role})
config.add_ec2_instance("CacheManager",
names.cache_manager,
aws.ami_lookup(session, "cachemanager.boss"),
keypair,
subnet=Ref("InternalSubnet"),
public_ip=False,
type_=const.CACHE_MANAGER_TYPE,
security_groups=[Ref("InternalSecurityGroup")],
user_data=parsed_user_data,
role="cachemanager")
lambda_bucket = aws.get_lambda_s3_bucket(session)
config.add_lambda("MultiLambda",
names.multi_lambda,
Ref("LambdaCacheExecutionRole"),
s3=(aws.get_lambda_s3_bucket(session),
"multilambda.{}.zip".format(domain),
"lambda_loader.handler"),
timeout=120,
memory=1024,
security_groups=[Ref('InternalSecurityGroup')],
subnets=lambda_subnets,
runtime='python3.6')
if creating_tile_bucket:
config.add_lambda_permission(
'tileBucketInvokeMultiLambda',
names.multi_lambda,
principal='s3.amazonaws.com',
source={
'Fn::Join':
[':', ['arn', 'aws', 's3', '', '', tile_bucket_name]]
}, #DP TODO: move into constants
depends_on=['tileBucket', 'MultiLambda'])
else:
config.add_lambda_permission(
'tileBucketInvokeMultiLambda',
names.multi_lambda,
principal='s3.amazonaws.com',
source={
'Fn::Join':
[':', ['arn', 'aws', 's3', '', '', tile_bucket_name]]
},
depends_on='MultiLambda')
# Add topic to indicating that the object store has been write locked.
# Now using "production mailing list" instead of separate write lock topic.
#config.add_sns_topic('WriteLock',
# names.write_lock_topic,
# names.write_lock,
# []) # TODO: add subscribers
return config
def create_config(session, domain, keypair=None, user_data=None, db_config={}):
"""Create the CloudFormationConfiguration object."""
names = AWSNames(domain)
config = CloudFormationConfiguration('proofreader', domain, const.REGION)
vpc_id = config.find_vpc(session)
az_subnets, _ = config.find_all_availability_zones(session)
external_subnet_id = aws.subnet_id_lookup(session,
names.subnet("external"))
config.add_arg(
Arg.Subnet("ExternalSubnet", external_subnet_id,
"ID of External Subnet to create resources in"))
sgs = aws.sg_lookup_all(session, vpc_id)
# Only allow unsecure web access from APL, until a ELB is configured for HTTPS
config.add_security_group("HttpSecurityGroup", names.http,
[("tcp", "80", "80", const.INCOMING_SUBNET)])
config.add_ec2_instance("ProofreaderWeb",
names.proofreader,
aws.ami_lookup(session, "proofreader-web.boss"),
keypair,
public_ip=True,
subnet=Ref("ExternalSubnet"),
security_groups=[
sgs[names.internal], sgs[names.ssh],
Ref('HttpSecurityGroup')
],
user_data=user_data,
depends_on="ProofreaderDB"
) # make sure the DB is launched before we start
config.add_rds_db("ProofreaderDB",
names.proofreader_db,
db_config.get("port"),
db_config.get("name"),
db_config.get("user"),
db_config.get("password"),
az_subnets,
security_groups=[sgs[names.internal]])
return config
def delete(bosslet_config):
config = CloudFormationConfiguration('copycuboid', bosslet_config)
config.delete()