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(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 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 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 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, db_config={}):
names = bosslet_config.names
session = bosslet_config.session
# 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.sns)
if dns_arn is None:
raise MissingResourceError('SNS topic', names.dns.sns)
mailing_list_arn = aws.sns_topic_lookup(session,
bosslet_config.ALERT_TOPIC)
if mailing_list_arn is None:
raise MissingResourceError('SNS topic', bosslet_config.ALERT_TOPIC)
# 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.dns
user_data["system"]["type"] = "endpoint"
user_data["aws"]["db"] = names.endpoint_db.rds
user_data["aws"]["cache"] = names.cache.redis
user_data["aws"]["cache-state"] = names.cache_state.redis
if const.REDIS_SESSION_TYPE is not None:
user_data["aws"]["cache-session"] = names.cache_session.redis
else:
# Don't create a Redis server for dev stacks.
user_data["aws"]["cache-session"] = ''
if const.REDIS_THROTTLE_TYPE is not None:
user_data["aws"]["cache-throttle"] = names.cache_throttle.redis
else:
user_data["aws"]["cache-throttle"] = ''
## 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"]["cache-throttle-db"] = "0"
user_data["aws"]["cache-session-db"] = "0"
user_data["aws"]["meta-db"] = names.meta.ddb
# 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.sqs)) # str(Ref("S3FlushQueue")) DP XXX
user_data["aws"]["s3-flush-deadletter-queue"] = str(
Ref(names.deadletter.sqs)) #str(Ref("DeadLetterQueue")) DP XXX
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"]["prod_mailing_list"] = mailing_list_arn
user_data["aws"]["max_task_id_suffix"] = str(const.MAX_TASK_ID_SUFFIX)
user_data["aws"]["id-index-new-chunk-threshold"] = str(
const.DYNAMO_ID_INDEX_NEW_CHUNK_THRESHOLD)
user_data["aws"]["index-deadletter-queue"] = str(
Ref(names.index_deadletter.sqs))
user_data["aws"]["index-cuboids-keys-queue"] = str(
Ref(names.index_cuboids_keys.sqs))
user_data["auth"]["OIDC_VERIFY_SSL"] = str(bosslet_config.VERIFY_SSL)
user_data["lambda"]["flush_function"] = names.multi_lambda.lambda_
user_data["lambda"]["page_in_function"] = names.multi_lambda.lambda_
user_data["lambda"]["ingest_function"] = names.tile_ingest.lambda_
user_data["lambda"]["downsample_volume"] = names.downsample_volume.lambda_
user_data["lambda"]["tile_uploaded_function"] = names.tile_uploaded.lambda_
user_data['sfn']['populate_upload_queue'] = names.ingest_queue_populate.sfn
user_data['sfn']['upload_sfn'] = names.ingest_queue_upload.sfn
user_data['sfn'][
'volumetric_upload_sfn'] = names.volumetric_ingest_queue_upload.sfn
user_data['sfn']['downsample_sfn'] = names.resolution_hierarchy.sfn
user_data['sfn'][
'index_cuboid_supervisor_sfn'] = names.index_cuboid_supervisor.sfn
# Prepare user data for parsing by CloudFormation.
parsed_user_data = {
"Fn::Join": ["", user_data.format_for_cloudformation()]
}
config = CloudFormationConfiguration('api', bosslet_config, version="2")
keypair = bosslet_config.SSH_KEY
vpc_id = config.find_vpc()
internal_subnets, external_subnets = config.find_all_subnets()
az_subnets_asg, external_subnets_asg = config.find_all_subnets(
compatibility='asg')
sgs = aws.sg_lookup_all(session, vpc_id)
# DP XXX: hack until we can get productio updated correctly
config.add_security_group(
'AllHttpHttpsSecurityGroup', names.https.sg,
[('tcp', '443', '443', bosslet_config.HTTPS_INBOUND),
('tcp', '80', '80', bosslet_config.HTTPS_INBOUND)])
sgs[names.https.sg] = Ref('AllHttpHttpsSecurityGroup')
# Create SQS queues and apply access control policies.
# Deadletter queue for indexing operations. This one is populated
# manually by states in the indexing step functions.
config.add_sqs_queue(names.index_deadletter.sqs,
names.index_deadletter.sqs, 30, 20160)
# Queue that holds S3 object keys of cuboids to be indexed.
config.add_sqs_queue(names.index_cuboids_keys.sqs,
names.index_cuboids_keys.sqs, 120, 20160)
#config.add_sqs_queue("DeadLetterQueue", names.deadletter.sqs, 30, 20160) DP XXX
config.add_sqs_queue(names.deadletter.sqs, names.deadletter.sqs, 30, 20160)
max_receives = 3
#config.add_sqs_queue("S3FlushQueue", DP XXX
config.add_sqs_queue(names.s3flush.sqs,
names.s3flush.sqs,
30,
dead=(Arn(names.deadletter.sqs), max_receives))
config.add_sqs_policy(
"sqsEndpointPolicy",
'sqsEndpointPolicy', # DP XXX
[Ref(names.deadletter.sqs),
Ref(names.s3flush.sqs)],
endpoint_role_arn)
config.add_sqs_policy(
"sqsCachemgrPolicy",
'sqsCachemgrPolicy', # DP XXX
[Ref(names.deadletter.sqs),
Ref(names.s3flush.sqs)],
cachemanager_role_arn)
# Create the endpoint ASG, ELB, and RDS instance
cert = aws.cert_arn_lookup(session, names.public_dns("api"))
target_group_keys = config.add_app_loadbalancer(
"EndpointAppLoadBalancer",
names.endpoint_elb.dns, [("443", "80", "HTTPS", cert)],
vpc_id=vpc_id,
subnets=external_subnets_asg,
security_groups=[sgs[names.internal.sg], sgs[names.https.sg]],
public=True)
target_group_arns = [Ref(key) for key in target_group_keys]
config.add_public_dns('EndpointAppLoadBalancer', names.public_dns('api'))
config.add_autoscale_group("Endpoint",
names.endpoint.dns,
aws.ami_lookup(bosslet_config,
names.endpoint.ami),
keypair,
subnets=az_subnets_asg,
type_=const.ENDPOINT_TYPE,
security_groups=[sgs[names.internal.sg]],
user_data=parsed_user_data,
min=const.ENDPOINT_CLUSTER_MIN,
max=const.ENDPOINT_CLUSTER_MAX,
notifications=dns_arn,
role=aws.instance_profile_arn_lookup(
session, 'endpoint'),
health_check_grace_period=90,
detailed_monitoring=True,
target_group_arns=target_group_arns,
depends_on=["EndpointDB"])
# 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.dns,
db_config.get("port"),
db_config.get("name"),
db_config.get("user"),
db_config.get("password"),
internal_subnets,
type_=const.RDS_TYPE,
security_groups=[sgs[names.internal.sg]])
# 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.ddb,
**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.ddb,
**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.ddb,
**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.ddb,
**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.ddb,
**dynamo_id_count_cfg) # DP XXX
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