-
Notifications
You must be signed in to change notification settings - Fork 13
/
client.py
766 lines (661 loc) · 28 KB
/
client.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
import json
import logging
import time
from typing import List, Optional
import boto3
import botocore
import dagster._check as check
from botocore.config import Config
from botocore.exceptions import ClientError
from dagster._utils.cached_method import cached_method
from dagster_aws.ecs.tasks import DagsterEcsTaskDefinitionConfig
from dagster_aws.ecs.utils import task_definitions_match
from dagster_cloud.workspace.ecs.service import Service
DEFAULT_ECS_TIMEOUT = 300
DEFAULT_ECS_GRACE_PERIOD = 30
STOPPED_TASK_GRACE_PERIOD = 30
ECS_EXEC_LINUX_PARAMETERS = {
"capabilities": {"add": ["SYS_PTRACE"]},
"initProcessEnabled": True,
}
config = Config(retries={"max_attempts": 50, "mode": "standard"})
def get_debug_ecs_prompt(cluster: str, task_arn: str) -> str:
return (
"For more information about the failure, check the ECS console for logs for task"
f" {task_arn} in cluster {cluster}."
)
class EcsServiceError(Exception):
def __init__(self, cluster, task_arn, stopped_reason, logs, show_debug_prompt: bool):
log_str = "Task logs:\n" + "\n".join(logs) if logs else "No task logs."
message = (
f"ECS service failed because task {task_arn} failed: {stopped_reason}\n\n{log_str}"
)
if show_debug_prompt:
message = message + f"\n\n{get_debug_ecs_prompt(cluster, task_arn)}"
super().__init__(message)
class Client:
def __init__(
self,
cluster_name: str,
service_discovery_namespace_id: str,
log_group: str,
subnet_ids: Optional[List[str]] = None,
security_group_ids: Optional[List[str]] = None,
ecs_client=None,
timeout: int = DEFAULT_ECS_TIMEOUT,
grace_period: int = DEFAULT_ECS_GRACE_PERIOD,
launch_type: str = "FARGATE",
show_debug_cluster_info: bool = True,
):
self.ecs = ecs_client if ecs_client else boto3.client("ecs", config=config)
self.logs = boto3.client("logs", config=config)
self.service_discovery = boto3.client("servicediscovery", config=config)
self.tags_client = boto3.client("resourcegroupstaggingapi", config=config)
self._use_legacy_tag_filtering = False
self.cluster_name = cluster_name.split("/")[-1]
self.subnet_ids = check.opt_list_param(subnet_ids, "subnet_ids")
self.security_group_ids = check.opt_list_param(security_group_ids, "security_group_ids")
self.service_discovery_namespace_id = check.str_param(
service_discovery_namespace_id, "service_discovery_namespace_id"
)
self.log_group = check.str_param(log_group, "log_group")
self.show_debug_cluster_info = show_debug_cluster_info
self.timeout = check.int_param(timeout, "timeout")
self.grace_period = check.int_param(grace_period, "grace_period")
self.launch_type = check.str_param(launch_type, "launch_type")
self._namespace: Optional[str] = None
@property
def ec2(self):
# Reconstruct the resource on each call because it's not threadsafe
# https://boto3.amazonaws.com/v1/documentation/api/latest/guide/resources.html#multithreading-or-multiprocessing-with-resources
return boto3.resource("ec2", config=config)
@property
def namespace(self):
if not self._namespace:
self._namespace = (
self.service_discovery.get_namespace(
Id=self.service_discovery_namespace_id,
)
.get("Namespace")
.get("Name")
)
return self._namespace
@property
def taggable(self):
settings = self.ecs.list_account_settings(
name="serviceLongArnFormat",
effectiveSettings=True,
)
return settings["settings"][0]["value"] == "enabled"
@property
@cached_method
def network_configuration(self):
network_configuration = {
"awsvpcConfiguration": {
"subnets": self.subnet_ids,
**(
{"assignPublicIp": self._assign_public_ip()}
if self.launch_type == "FARGATE"
else {}
),
},
}
if self.security_group_ids:
network_configuration["awsvpcConfiguration"]["securityGroups"] = self.security_group_ids
return network_configuration
def register_task_definition(
self,
family,
image,
command,
execution_role_arn,
container_name=None,
task_role_arn=None,
env=None,
secrets=None,
sidecars=None,
logger=None,
cpu=None,
memory=None,
ephemeral_storage=None,
repository_credentials=None,
runtime_platform=None,
mount_points=None,
volumes=None,
linux_parameters=None,
):
container_name = container_name or family
logger = logger or logging.getLogger("dagster_cloud.EcsClient")
env = env or {}
environment = [{"name": key, "value": value} for key, value in env.items()]
secrets = (
[{"name": key, "valueFrom": value} for key, value in secrets.items()] if secrets else []
)
task_definition_config = DagsterEcsTaskDefinitionConfig(
family=family,
image=image,
container_name=container_name,
command=command,
log_configuration={
"logDriver": "awslogs",
"options": {
"awslogs-group": self.log_group,
"awslogs-region": self.ecs.meta.region_name,
"awslogs-stream-prefix": family,
},
},
secrets=secrets,
environment=environment,
execution_role_arn=execution_role_arn,
task_role_arn=task_role_arn,
sidecars=sidecars,
requires_compatibilities=[self.launch_type],
cpu=cpu,
memory=memory,
ephemeral_storage=ephemeral_storage,
repository_credentials=repository_credentials,
runtime_platform=runtime_platform,
mount_points=mount_points,
volumes=volumes,
linux_parameters=linux_parameters,
)
try:
existing_task_definition = self.ecs.describe_task_definition(taskDefinition=family)[
"taskDefinition"
]
except ClientError:
logger.info("No existing task definition")
# task definition does not exist, do not reuse
existing_task_definition = None
if not existing_task_definition or not task_definitions_match(
task_definition_config,
existing_task_definition,
container_name=container_name,
):
task_definition_arn = (
self.ecs.register_task_definition(**task_definition_config.task_definition_dict())
.get("taskDefinition")
.get("taskDefinitionArn")
)
logger.info(f"Created new task definition {task_definition_arn}")
else:
task_definition_arn = existing_task_definition["taskDefinitionArn"]
logger.info(f"Re-using existing task definition {task_definition_arn}")
return task_definition_arn
def create_service(
self,
name,
image,
command,
execution_role_arn,
family=None,
container_name=None,
task_role_arn=None,
env=None,
tags=None,
register_service_discovery=True,
secrets=None,
sidecars=None,
logger=None,
cpu=None,
memory=None,
ephemeral_storage=None,
replica_count=None,
repository_credentials=None,
allow_ecs_exec=False,
runtime_platform=None,
mount_points=None,
volumes=None,
):
logger = logger or logging.getLogger("dagster_cloud.EcsClient")
service_name = name
family = family or name
logger.info(f"Checking if task definition {family} for {name} can be re-used...")
task_definition_arn = self.register_task_definition(
family=family,
image=image,
container_name=container_name,
execution_role_arn=execution_role_arn,
task_role_arn=task_role_arn,
command=command,
env=env or {},
secrets=secrets,
sidecars=sidecars,
logger=logger,
cpu=cpu,
memory=memory,
ephemeral_storage=ephemeral_storage,
repository_credentials=repository_credentials,
runtime_platform=runtime_platform,
mount_points=mount_points,
volumes=volumes,
linux_parameters=ECS_EXEC_LINUX_PARAMETERS if allow_ecs_exec else None,
)
service_registry_arn = None
# Configure service discovery
if register_service_discovery:
logger.info(f"Creating service registry for {service_name}...")
service_registry_arn = self._create_service_registry(
service_name=service_name,
tags=tags,
)
logger.info(f"Creating ECS service {service_name}...")
# Create the service
return self._create_service(
service_name=service_name,
service_registry_arn=service_registry_arn,
task_definition_arn=task_definition_arn,
replica_count=replica_count,
tags=tags,
allow_ecs_exec=allow_ecs_exec,
)
def delete_service(
self,
service,
logger=None,
):
logger = logger or logging.getLogger("dagster_cloud.EcsClient")
# Reduce running tasks to 0
try:
self.ecs.update_service(
cluster=self.cluster_name,
service=service.name,
desiredCount=0,
)
except botocore.exceptions.ClientError as error:
if error.response["Error"]["Code"] in [
"ServiceNotFoundException",
"ServiceNotActiveException",
]:
logger.warning(
f"Service {self.cluster_name}/{service.name} is not found or inactive: {error}"
)
return
raise
# Delete the ECS service
self.ecs.delete_service(
cluster=self.cluster_name,
service=service.name,
force=True,
)
# get service discovery id
service_discovery_id = self._get_service_discovery_id(
service.hostname,
)
if service_discovery_id:
# Unregister dangling ecs tasks from service discovery
instances_paginator = self.service_discovery.get_paginator("list_instances")
instances = instances_paginator.paginate(
ServiceId=service_discovery_id,
).build_full_result()["Instances"]
deregister_operation_ids = []
for instance in instances:
resp = self.service_discovery.deregister_instance(
ServiceId=service_discovery_id, InstanceId=instance["Id"]
)
deregister_operation_ids.append(resp["OperationId"])
# wait for instances to complete deregistering
for operation_id in deregister_operation_ids:
status = ""
while status != "SUCCESS":
status = self.service_discovery.get_operation(OperationId=operation_id)[
"Operation"
]["Status"]
if status == "FAIL":
raise Exception("deregister operation failed")
time.sleep(2)
# delete service discovery
self.service_discovery.delete_service(
Id=service_discovery_id,
)
def list_services(self, tags=None, logger=None):
logger = logger or logging.getLogger("dagster_cloud.EcsClient")
services = []
if tags:
if not self._use_legacy_tag_filtering:
try:
# obtain services present in the cluster
ecs_services_paginator = self.ecs.get_paginator("list_services")
actual_services = []
for page in ecs_services_paginator.paginate(cluster=self.cluster_name):
actual_services.extend(page["serviceArns"])
# obtain services with the specified tags
tag_paginator = self.tags_client.get_paginator("get_resources")
for page in tag_paginator.paginate(
TagFilters=[
{
"Key": key,
"Values": [value],
}
for key, value in tags.items()
],
ResourceTypeFilters=["ecs:service"],
):
for resource in page["ResourceTagMappingList"]:
resource_arn = resource["ResourceARN"]
# note: the Resource Group Tagging API may return resources corresponding to services
# that are no longer present in the cluster, so we filter out those that are not present
if resource_arn in actual_services:
services.append(Service(client=self, arn=resource_arn))
except botocore.exceptions.ClientError as error:
if error.response["Error"]["Code"] == "AccessDeniedException":
self._use_legacy_tag_filtering = True
logger.warning(
"dagster-cloud 1.4.0 includes performance enhancements that rely on the"
" Resource"
" Group Tagging API. To take advantage, add 'tags:GetResources'"
" permissions to"
" your"
" task role."
" https://docs.aws.amazon.com/resourcegroupstagging/latest/APIReference/overview.html"
)
services = self.list_services(tags)
else:
raise
else:
filtered_services = [
service
for service in self.list_services()
if tags.items() <= service.tags.items()
]
return filtered_services
else:
service_paginator = self.ecs.get_paginator("list_services")
for page in service_paginator.paginate(cluster=self.cluster_name):
for service_arn in page.get("serviceArns"):
service = Service(client=self, arn=service_arn)
services.append(service)
return services
def _run_task(self, name, image, command, execution_role_arn):
"""This is no longer used except as an integration test util
for testing various network configurations.
"""
task_definition_arn = self.register_task_definition(
family=name,
image=image,
command=command,
execution_role_arn=execution_role_arn,
)
task_arn = (
self.ecs.run_task(
taskDefinition=task_definition_arn,
cluster=self.cluster_name,
launchType=self.launch_type,
networkConfiguration=self.network_configuration,
)
.get("tasks", [{}])[0]
.get("taskArn")
)
self.ecs.get_waiter("tasks_stopped").wait(
cluster=self.cluster_name,
tasks=[task_arn],
WaiterConfig={"Delay": 1, "MaxAttempts": self.timeout},
)
exit_code = (
self.ecs.describe_tasks(
cluster=self.cluster_name,
tasks=[task_arn],
)
.get("tasks", [{}])[0]
.get("containers", [{}])[0]
.get("exitCode")
)
if exit_code:
raise Exception(self.get_task_logs(task_arn))
return True
def _create_service_registry(self, service_name, tags=None):
if not tags:
tags = {}
# https://docs.aws.amazon.com/AmazonECS/latest/developerguide/create-service-discovery.html
service_registry_arn = (
self.service_discovery.create_service(
Name=service_name,
NamespaceId=self.service_discovery_namespace_id,
DnsConfig={
"DnsRecords": [
{"Type": "A", "TTL": 60},
]
},
Tags=[{"Key": key, "Value": value} for key, value in tags.items()],
)
.get("Service", {})
.get("Arn")
)
return service_registry_arn
def _create_service(
self,
service_name,
task_definition_arn,
service_registry_arn,
replica_count=None,
tags=None,
allow_ecs_exec=False,
):
params = dict(
cluster=self.cluster_name,
serviceName=service_name,
taskDefinition=task_definition_arn,
launchType=self.launch_type,
desiredCount=replica_count or 1,
enableExecuteCommand=allow_ecs_exec,
propagateTags="SERVICE",
)
params["networkConfiguration"] = self.network_configuration
if service_registry_arn:
params["serviceRegistries"] = [{"registryArn": service_registry_arn}]
if tags and self.taggable:
params["tags"] = [
{"key": key, **({"value": value} if value is not None else {})}
for key, value in tags.items()
]
arn = self.ecs.create_service(**params).get("service").get("serviceArn")
return Service(client=self, arn=arn)
def wait_for_new_service(self, service, container_name, logger=None) -> dict:
logger = logger or logging.getLogger("dagster_cloud.EcsClient")
service_name = service.name
start_time = time.time()
while True:
response = self.ecs.describe_services(
cluster=self.cluster_name,
services=[service_name],
)
if response.get("services"):
running_tasks = self.check_service_has_running_tasks(
service_name, container_name, logger=logger
)
return running_tasks[0]
failures = response.get("failures")
if not failures:
# This might not be a possible state; it's unclear if the ECS API can return empty lists for both
# "failures" and "services":
# https://docs.aws.amazon.com/AmazonECS/latest/APIReference/API_DescribeServices.html
raise Exception(
"ECS DescribeServices API returned an empty response for service"
f" {service.arn}."
)
# Even if we fail, check a few more times in case it's just the ECS API
# being eventually consistent
if time.time() - start_time >= self.grace_period:
raise Exception(
f"ECS DescribeServices API returned failures: {json.dumps(failures)}"
)
time.sleep(10)
def _raise_failed_task(self, task, container_name, logger):
task_arn = task["taskArn"]
stopped_reason = task["stoppedReason"]
logs = None
try:
logs = self.get_task_logs(task_arn, container_name=container_name)
except:
logger.exception(f"Error trying to get logs for failed task {task_arn}")
raise EcsServiceError(
cluster=self.cluster_name,
task_arn=task_arn,
stopped_reason=stopped_reason,
logs=logs,
show_debug_prompt=self.show_debug_cluster_info,
)
def assert_task_not_stopped(self, task_arn, container_name, logger=None):
logger = logger or logging.getLogger("dagster_cloud.EcsClient")
task = self.ecs.describe_tasks(cluster=self.cluster_name, tasks=[task_arn]).get("tasks")[0]
if task.get("lastStatus") == "STOPPED":
self._raise_failed_task(task, container_name, logger)
def check_service_has_running_tasks(
self, service_name, container_name, logger=None
) -> List[dict]:
# return the ARN of the task if it starts
logger = logger or logging.getLogger("dagster_cloud.EcsClient")
start_time = time.time()
tasks_to_track = None
while start_time + self.timeout > time.time():
# Check for running tasks to track
if not tasks_to_track:
services = self.ecs.describe_services(
cluster=self.cluster_name,
services=[service_name],
)
if not services:
raise Exception(
f"Service description not found for {self.cluster_name}/{service_name}"
)
service = services.get("services")[0]
desired_count = service.get("desiredCount")
running_count = service.get("runningCount")
# If the service has reached the desired count, we can start tracking the tasks
if desired_count == running_count:
running_tasks = self.ecs.list_tasks(
cluster=self.cluster_name,
serviceName=service_name,
desiredStatus="RUNNING",
).get("taskArns")
if not running_tasks:
raise Exception(
f"Unexpected error obtaining tasks for {service_name} in {self.cluster_name}"
)
tasks_to_track = running_tasks
elif time.time() > start_time + STOPPED_TASK_GRACE_PERIOD:
# If there are still no running_tasks tasks after a certain grace period, check for stopped tasks
stopped_tasks = self._check_for_stopped_tasks(service_name)
if stopped_tasks:
self._raise_failed_task(stopped_tasks[0], container_name, logger)
if tasks_to_track:
tasks = self.ecs.describe_tasks(
cluster=self.cluster_name, tasks=tasks_to_track
).get("tasks")
# Wait for all tasks essential containers to be running
all_tasks_running = len(tasks) > 0
for task in tasks:
if task.get("lastStatus") == "RUNNING":
if not self._check_all_essential_containers_are_running(task):
all_tasks_running = False
elif task.get("lastStatus") == "STOPPED":
self._raise_failed_task(task, container_name, logger)
if all_tasks_running:
return tasks_to_track
time.sleep(20)
# Fetch the service event logs to try to get some clue about why the service never spun
# up any tasks
service_events_str = ""
try:
response = self.ecs.describe_services(
cluster=self.cluster_name,
services=[service_name],
)
if response.get("services"):
service = response["services"][0]
service_events = [event.get("message") for event in service.get("events")]
service_events_str = "Service events:\n" + "\n".join(service_events)
except:
logger.exception(f"Error trying to get service event logs from service {service_name}")
raise Exception(
f"Timed out waiting for a running task for service: {service_name}."
f" {service_events_str}"
)
def _check_for_stopped_tasks(self, service_name):
stopped = self.ecs.list_tasks(
cluster=self.cluster_name,
serviceName=service_name,
desiredStatus="STOPPED",
).get("taskArns")
if stopped:
stopped_tasks = self.ecs.describe_tasks(cluster=self.cluster_name, tasks=stopped).get(
"tasks"
)
stopped_tasks = sorted(
stopped_tasks,
key=lambda task: task["createdAt"].timestamp(),
reverse=True,
)
return stopped_tasks
return []
def _check_all_essential_containers_are_running(self, task) -> bool:
task_definition = self.ecs.describe_task_definition(
taskDefinition=task.get("taskDefinitionArn"),
).get("taskDefinition")
essential_containers = {
container["name"]
for container in task_definition["containerDefinitions"]
if container["essential"]
}
# Just because the task is RUNNING doesn't mean everything has started up correctly -
# sometimes it briefly thinks it is RUNNING even though individual containers have STOPPED.
# Wait for all essential containers to be running_tasks too before declaring victory.
return all(
container["name"] not in essential_containers or container["lastStatus"] == "RUNNING"
for container in task["containers"]
)
def _get_service_discovery_id(self, hostname):
service_name = hostname.split("." + self.namespace)[0]
paginator = self.service_discovery.get_paginator("list_services")
for page in paginator.paginate(
Filters=[
{
"Name": "NAMESPACE_ID",
"Values": [
self.service_discovery_namespace_id,
],
"Condition": "EQ",
},
],
):
for service in page["Services"]:
if service["Name"] == service_name:
return service["Id"]
def _assign_public_ip(self):
# https://docs.aws.amazon.com/AmazonECS/latest/userguide/fargate-task-networking.html
# Assign a public IP if any of the subnets are public
route_tables = self.ec2.route_tables.filter(
Filters=[
{"Name": "association.subnet-id", "Values": self.subnet_ids},
]
)
# Consider a subnet to be public if it has a route that targets
# an internet gateway; private subnets have routes that target NAT gateways
for route_table in route_tables:
if any(route.nat_gateway_id for route in route_table.routes):
return "DISABLED"
return "ENABLED"
def get_task_logs(self, task_arn, container_name, limit=10):
task = self.ecs.describe_tasks(cluster=self.cluster_name, tasks=[task_arn]).get("tasks")[0]
task_definition_arn = task.get("taskDefinitionArn")
task_definition = self.ecs.describe_task_definition(taskDefinition=task_definition_arn).get(
"taskDefinition"
)
matching_container_definitions = [
container_definition
for container_definition in task_definition.get("containerDefinitions", [])
if container_definition["name"] == container_name
]
if not matching_container_definitions:
raise Exception(f"Could not find container with name {container_name}")
container_definition = matching_container_definitions[0]
log_stream_prefix = (
container_definition.get("logConfiguration").get("options").get("awslogs-stream-prefix")
)
container_name = container_definition.get("name")
task_id = task_arn.split("/")[-1]
log_stream = f"{log_stream_prefix}/{container_name}/{task_id}"
events = self.logs.get_log_events(
logGroupName=self.log_group, logStreamName=log_stream, limit=limit
).get("events")
return [event.get("message") for event in events]