def scan_scan_unit(scan_unit: ScanUnit) -> Tuple[str, Dict[str, Any]]:
logger = Logger()
with logger.bind(
account_id=scan_unit.account_id,
region=scan_unit.region_name,
service=scan_unit.service,
resource_classes=sorted([
resource_spec_class.__name__
for resource_spec_class in scan_unit.resource_spec_classes
]),
):
start_t = time.time()
logger.info(event=AWSLogEvents.ScanAWSAccountServiceStart)
session = boto3.Session(
aws_access_key_id=scan_unit.access_key,
aws_secret_access_key=scan_unit.secret_key,
aws_session_token=scan_unit.token,
region_name=scan_unit.region_name,
)
scan_accessor = AWSAccessor(session=session,
account_id=scan_unit.account_id,
region_name=scan_unit.region_name)
graph_spec = GraphSpec(
name=scan_unit.graph_name,
version=scan_unit.graph_version,
resource_spec_classes=scan_unit.resource_spec_classes,
scan_accessor=scan_accessor,
)
start_time = int(time.time())
resources: List[Resource] = []
errors = []
try:
resources = graph_spec.scan()
except Exception as ex:
error_str = str(ex)
trace_back = traceback.format_exc()
logger.error(event=AWSLogEvents.ScanAWSAccountError,
error=error_str,
trace_back=trace_back)
error = f"{str(ex)}\n{trace_back}"
errors.append(error)
end_time = int(time.time())
graph_set = GraphSet(
name=scan_unit.graph_name,
version=scan_unit.graph_version,
start_time=start_time,
end_time=end_time,
resources=resources,
errors=errors,
stats=scan_accessor.api_call_stats,
)
end_t = time.time()
elapsed_sec = end_t - start_t
logger.info(event=AWSLogEvents.ScanAWSAccountServiceEnd,
elapsed_sec=elapsed_sec)
return (scan_unit.account_id, graph_set.to_dict())
class TestGraphSetWithValidDataNoMerging(TestCase):
def setUp(self):
resource_a1 = Resource(resource_id="123",
type_name="test:a",
links=[SimpleLink(pred="has-foo", obj="goo")])
resource_a2 = Resource(resource_id="456", type_name="test:a")
resource_b1 = Resource(
resource_id="abc",
type_name="test:b",
links=[ResourceLinkLink(pred="has-a", obj="123")])
resource_b2 = Resource(resource_id="def",
type_name="test:b",
links=[SimpleLink(pred="name", obj="sue")])
resources = [resource_a1, resource_a2, resource_b1, resource_b2]
self.graph_set = GraphSet(
name="test-name",
version="1",
start_time=1234,
end_time=4567,
resources=resources,
errors=["test err 1", "test err 2"],
stats=MultilevelCounter(),
)
def test_rdf_a_type(self):
graph = self.graph_set.to_rdf()
a_results = graph.query(
"select ?p ?o where {?s a <test-name:test:a> ; ?p ?o} order by ?p ?o"
)
expected_a_result_tuples = [
("http://www.w3.org/1999/02/22-rdf-syntax-ns#type",
"test-name:test:a"),
("http://www.w3.org/1999/02/22-rdf-syntax-ns#type",
"test-name:test:a"),
("test-name:has-foo", "goo"),
("test-name:id", "123"),
("test-name:id", "456"),
]
a_result_tuples = []
for a_result in a_results:
self.assertEqual(2, len(a_result))
a_result_tuples.append((str(a_result[0]), str(a_result[1])))
self.assertEqual(expected_a_result_tuples, a_result_tuples)
def test_rdf_b_type(self):
graph = self.graph_set.to_rdf()
graph.serialize("/tmp/test.rdf")
linked_a_node_results = graph.query(
"select ?s where {?s a <test-name:test:a>; <test-name:id> '123' }")
self.assertEqual(len(linked_a_node_results), 1)
for linked_a_node_result in linked_a_node_results:
linked_a_node = str(linked_a_node_result[0])
b_results = graph.query(
"select ?p ?o where {?s a <test-name:test:b> ; ?p ?o} order by ?p ?o"
)
expected_b_result_tuples = [
("http://www.w3.org/1999/02/22-rdf-syntax-ns#type",
"test-name:test:b"),
("http://www.w3.org/1999/02/22-rdf-syntax-ns#type",
"test-name:test:b"),
("test-name:has-a", str(linked_a_node)),
("test-name:id", "abc"),
("test-name:id", "def"),
("test-name:name", "sue"),
]
b_result_tuples = []
for b_result in b_results:
self.assertEqual(2, len(b_result))
b_result_tuples.append((str(b_result[0]), str(b_result[1])))
self.assertEqual(expected_b_result_tuples, b_result_tuples)
def test_rdf_error_graphing(self):
graph = self.graph_set.to_rdf()
err_results = graph.query(
"select ?o where { ?s <test-name:error> ?o } order by ?o")
err_strs = []
expected_err_strs = ["test err 1", "test err 2"]
for err_result in err_results:
self.assertEqual(1, len(err_result))
err_strs.append(str(err_result[0]))
self.assertEqual(err_strs, expected_err_strs)
def test_to_dict(self):
expected_dict = {
"name": "test-name",
"version": "1",
"start_time": 1234,
"end_time": 4567,
"resources": {
"123": {
"type": "test:a",
"links": [{
"pred": "has-foo",
"obj": "goo",
"type": "simple"
}],
},
"456": {
"type": "test:a"
},
"abc": {
"type":
"test:b",
"links": [{
"pred": "has-a",
"obj": "123",
"type": "resource_link"
}],
},
"def": {
"type": "test:b",
"links": [{
"pred": "name",
"obj": "sue",
"type": "simple"
}],
},
},
"errors": ["test err 1", "test err 2"],
"stats": {
"count": 0
},
}
self.assertDictEqual(expected_dict, self.graph_set.to_dict())
def test_from_dict(self):
input_dict = {
"name": "test-name",
"version": "1",
"start_time": 1234,
"end_time": 4567,
"resources": {
"123": {
"type": "test:a",
"links": [{
"pred": "has-foo",
"obj": "goo",
"type": "simple"
}],
},
"456": {
"type": "test:a"
},
"abc": {
"type":
"test:b",
"links": [{
"pred": "has-a",
"obj": "123",
"type": "resource_link"
}],
},
"def": {
"type": "test:b",
"links": [{
"pred": "name",
"obj": "sue",
"type": "simple"
}],
},
},
"errors": ["test err 1", "test err 2"],
"stats": {
"count": 0
},
}
graph_set = GraphSet.from_dict(input_dict)
self.assertEqual(graph_set.to_dict(), input_dict)
def test_validate(self):
self.graph_set.validate()
def scan(self) -> Dict[str, Any]:
"""Scan an account and return a dict containing keys:
* account_id: str
* output_artifact: str
* api_call_stats: Dict[str, Any]
* errors: List[str]
If errors is non-empty the results are incomplete for this account.
output_artifact is a pointer to the actual scan data - either on the local fs or in s3.
To scan an account we create a set of GraphSpecs, one for each region. Any ACCOUNT
level granularity resources are only scanned in a single region (e.g. IAM Users)
Returns:
Dict of scan result, see above for details.
"""
logger = Logger()
with logger.bind(account_id=self.account_id):
logger.info(event=AWSLogEvents.ScanAWSAccountStart)
output_artifact = None
stats = MultilevelCounter()
errors: List[str] = []
now = int(time.time())
try:
account_graph_set = GraphSet(
name=self.graph_name,
version=self.graph_version,
start_time=now,
end_time=now,
resources=[],
errors=[],
stats=stats,
)
# sanity check
session = self.get_session()
sts_client = session.client("sts")
sts_account_id = sts_client.get_caller_identity()["Account"]
if sts_account_id != self.account_id:
raise ValueError(
f"BUG: sts detected account_id {sts_account_id} != {self.account_id}"
)
if self.regions:
scan_regions = tuple(self.regions)
else:
scan_regions = get_all_enabled_regions(session=session)
# build graph specs.
# build a dict of regions -> services -> List[AWSResourceSpec]
regions_services_resource_spec_classes: DefaultDict[
str,
DefaultDict[str,
List[Type[AWSResourceSpec]]]] = defaultdict(
lambda: defaultdict(list))
resource_spec_class: Type[AWSResourceSpec]
for resource_spec_class in self.resource_spec_classes:
client_name = resource_spec_class.get_client_name()
resource_class_scan_granularity = resource_spec_class.scan_granularity
if resource_class_scan_granularity == ScanGranularity.ACCOUNT:
regions_services_resource_spec_classes[scan_regions[
0]][client_name].append(resource_spec_class)
elif resource_class_scan_granularity == ScanGranularity.REGION:
for region in scan_regions:
regions_services_resource_spec_classes[region][
client_name].append(resource_spec_class)
else:
raise NotImplementedError(
f"ScanGranularity {resource_class_scan_granularity} not implemented"
)
with ThreadPoolExecutor(
max_workers=self.max_svc_threads) as executor:
futures = []
for (
region,
services_resource_spec_classes,
) in regions_services_resource_spec_classes.items():
for (
service,
resource_spec_classes,
) in services_resource_spec_classes.items():
region_session = self.get_session(region=region)
region_creds = region_session.get_credentials()
scan_future = schedule_scan_services(
executor=executor,
graph_name=self.graph_name,
graph_version=self.graph_version,
account_id=self.account_id,
region=region,
service=service,
access_key=region_creds.access_key,
secret_key=region_creds.secret_key,
token=region_creds.token,
resource_spec_classes=tuple(
resource_spec_classes),
)
futures.append(scan_future)
for future in as_completed(futures):
graph_set_dict = future.result()
graph_set = GraphSet.from_dict(graph_set_dict)
errors += graph_set.errors
account_graph_set.merge(graph_set)
account_graph_set.validate()
except Exception as ex:
error_str = str(ex)
trace_back = traceback.format_exc()
logger.error(event=AWSLogEvents.ScanAWSAccountError,
error=error_str,
trace_back=trace_back)
errors.append(" : ".join((error_str, trace_back)))
unscanned_account_resource = UnscannedAccountResourceSpec.create_resource(
account_id=self.account_id, errors=errors)
account_graph_set = GraphSet(
name=self.graph_name,
version=self.graph_version,
start_time=now,
end_time=now,
resources=[unscanned_account_resource],
errors=errors,
stats=stats,
)
account_graph_set.validate()
output_artifact = self.artifact_writer.write_artifact(
name=self.account_id, data=account_graph_set.to_dict())
logger.info(event=AWSLogEvents.ScanAWSAccountEnd)
api_call_stats = account_graph_set.stats.to_dict()
return {
"account_id": self.account_id,
"output_artifact": output_artifact,
"errors": errors,
"api_call_stats": api_call_stats,
}
def scan(self) -> List[Dict[str, Any]]:
logger = Logger()
scan_result_dicts = []
now = int(time.time())
prescan_account_ids_errors: DefaultDict[str,
List[str]] = defaultdict(list)
futures = []
with ThreadPoolExecutor(max_workers=self.max_threads) as executor:
shuffled_account_ids = random.sample(
self.account_scan_plan.account_ids,
k=len(self.account_scan_plan.account_ids))
for account_id in shuffled_account_ids:
with logger.bind(account_id=account_id):
logger.info(event=AWSLogEvents.ScanAWSAccountStart)
try:
session = self.account_scan_plan.accessor.get_session(
account_id=account_id)
# sanity check
sts_client = session.client("sts")
sts_account_id = sts_client.get_caller_identity(
)["Account"]
if sts_account_id != account_id:
raise ValueError(
f"BUG: sts detected account_id {sts_account_id} != {account_id}"
)
if self.account_scan_plan.regions:
scan_regions = tuple(
self.account_scan_plan.regions)
else:
scan_regions = get_all_enabled_regions(
session=session)
account_gran_scan_region = random.choice(
self.preferred_account_scan_regions)
# build a dict of regions -> services -> List[AWSResourceSpec]
regions_services_resource_spec_classes: DefaultDict[
str, DefaultDict[
str,
List[Type[AWSResourceSpec]]]] = defaultdict(
lambda: defaultdict(list))
resource_spec_class: Type[AWSResourceSpec]
for resource_spec_class in self.resource_spec_classes:
client_name = resource_spec_class.get_client_name()
if resource_spec_class.scan_granularity == ScanGranularity.ACCOUNT:
if resource_spec_class.region_whitelist:
account_resource_scan_region = resource_spec_class.region_whitelist[
0]
else:
account_resource_scan_region = account_gran_scan_region
regions_services_resource_spec_classes[
account_resource_scan_region][
client_name].append(
resource_spec_class)
elif resource_spec_class.scan_granularity == ScanGranularity.REGION:
if resource_spec_class.region_whitelist:
resource_scan_regions = tuple(
region for region in scan_regions
if region in
resource_spec_class.region_whitelist)
if not resource_scan_regions:
resource_scan_regions = resource_spec_class.region_whitelist
else:
resource_scan_regions = scan_regions
for region in resource_scan_regions:
regions_services_resource_spec_classes[
region][client_name].append(
resource_spec_class)
else:
raise NotImplementedError(
f"ScanGranularity {resource_spec_class.scan_granularity} unimplemented"
)
# Build and submit ScanUnits
shuffed_regions_services_resource_spec_classes = random.sample(
regions_services_resource_spec_classes.items(),
len(regions_services_resource_spec_classes),
)
for (
region,
services_resource_spec_classes,
) in shuffed_regions_services_resource_spec_classes:
region_session = self.account_scan_plan.accessor.get_session(
account_id=account_id, region_name=region)
region_creds = region_session.get_credentials()
shuffled_services_resource_spec_classes = random.sample(
services_resource_spec_classes.items(),
len(services_resource_spec_classes),
)
for (
service,
svc_resource_spec_classes,
) in shuffled_services_resource_spec_classes:
future = schedule_scan(
executor=executor,
graph_name=self.graph_name,
graph_version=self.graph_version,
account_id=account_id,
region_name=region,
service=service,
access_key=region_creds.access_key,
secret_key=region_creds.secret_key,
token=region_creds.token,
resource_spec_classes=tuple(
svc_resource_spec_classes),
)
futures.append(future)
except Exception as ex:
error_str = str(ex)
trace_back = traceback.format_exc()
logger.error(
event=AWSLogEvents.ScanAWSAccountError,
error=error_str,
trace_back=trace_back,
)
prescan_account_ids_errors[account_id].append(
f"{error_str}\n{trace_back}")
account_ids_graph_set_dicts: Dict[str,
List[Dict[str,
Any]]] = defaultdict(list)
for future in as_completed(futures):
account_id, graph_set_dict = future.result()
account_ids_graph_set_dicts[account_id].append(graph_set_dict)
# first make sure no account id appears both in account_ids_graph_set_dicts
# and prescan_account_ids_errors - this should never happen
doubled_accounts = set(
account_ids_graph_set_dicts.keys()).intersection(
set(prescan_account_ids_errors.keys()))
if doubled_accounts:
raise Exception((
f"BUG: Account(s) {doubled_accounts} in both "
"account_ids_graph_set_dicts and prescan_account_ids_errors."))
# graph prescan error accounts
for account_id, errors in prescan_account_ids_errors.items():
with logger.bind(account_id=account_id):
unscanned_account_resource = UnscannedAccountResourceSpec.create_resource(
account_id=account_id, errors=errors)
account_graph_set = GraphSet(
name=self.graph_name,
version=self.graph_version,
start_time=now,
end_time=now,
resources=[unscanned_account_resource],
errors=errors,
stats=MultilevelCounter(),
)
account_graph_set.validate()
output_artifact = self.artifact_writer.write_json(
name=account_id, data=account_graph_set.to_dict())
logger.info(event=AWSLogEvents.ScanAWSAccountEnd)
api_call_stats = account_graph_set.stats.to_dict()
scan_result_dicts.append({
"account_id": account_id,
"output_artifact": output_artifact,
"errors": errors,
"api_call_stats": api_call_stats,
})
# graph rest
for account_id, graph_set_dicts in account_ids_graph_set_dicts.items():
with logger.bind(account_id=account_id):
# if there are any errors whatsoever we generate an empty graph with
# errors only
errors = []
for graph_set_dict in graph_set_dicts:
errors += graph_set_dict["errors"]
if errors:
unscanned_account_resource = UnscannedAccountResourceSpec.create_resource(
account_id=account_id, errors=errors)
account_graph_set = GraphSet(
name=self.graph_name,
version=self.graph_version,
start_time=now,
end_time=now,
resources=[unscanned_account_resource],
errors=errors,
stats=MultilevelCounter(
), # ENHANCHMENT: could technically get partial stats.
)
account_graph_set.validate()
else:
account_graph_set = GraphSet(
name=self.graph_name,
version=self.graph_version,
start_time=now,
end_time=now,
resources=[],
errors=[],
stats=MultilevelCounter(),
)
for graph_set_dict in graph_set_dicts:
graph_set = GraphSet.from_dict(graph_set_dict)
account_graph_set.merge(graph_set)
output_artifact = self.artifact_writer.write_json(
name=account_id, data=account_graph_set.to_dict())
logger.info(event=AWSLogEvents.ScanAWSAccountEnd)
api_call_stats = account_graph_set.stats.to_dict()
scan_result_dicts.append({
"account_id": account_id,
"output_artifact": output_artifact,
"errors": errors,
"api_call_stats": api_call_stats,
})
return scan_result_dicts