Skip to content

prefect_azure.container_instance

Integrations with the Azure Container Instances service. Note this module is experimental. The interfaces within may change without notice.

The AzureContainerInstanceJob infrastructure block in this module is ideally configured via the Prefect UI and run via a Prefect agent, but it can be called directly as demonstrated in the following examples.

Examples:

Run a command using an Azure Container Instances container.

AzureContainerInstanceJob(command=["echo", "hello world"]).run()

Run a command and stream the container's output to the local terminal.

AzureContainerInstanceJob(
    command=["echo", "hello world"],
    stream_output=True,
)

Run a command with a specific image

AzureContainerInstanceJob(command=["echo", "hello world"], image="alpine:latest")

Run a task with custom memory and CPU requirements

AzureContainerInstanceJob(command=["echo", "hello world"], memory=1.0, cpu=1.0)

Run a task with custom memory and CPU requirements

AzureContainerInstanceJob(command=["echo", "hello world"], memory=1.0, cpu=1.0)

Run a task with custom memory, CPU, and GPU requirements

AzureContainerInstanceJob(command=["echo", "hello world"], memory=1.0, cpu=1.0,
gpu_count=1, gpu_sku="V100")

Run a task with custom environment variables

AzureContainerInstanceJob(
    command=["echo", "hello $PLANET"],
    env={"PLANET": "earth"}
)

Classes

AzureContainerInstanceJob

Bases: Infrastructure

Run a command using a container on Azure Container Instances. Note this block is experimental. The interface may change without notice.

Source code in prefect_azure/container_instance.py
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
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
class AzureContainerInstanceJob(Infrastructure):
    """
    <span class="badge-api experimental"/>
    Run a command using a container on Azure Container Instances.
    Note this block is experimental. The interface may change without notice.
    """

    _block_type_name = "Azure Container Instance Job"
    _logo_url = "https://images.ctfassets.net/gm98wzqotmnx/6AiQ6HRIft8TspZH7AfyZg/39fd82bdbb186db85560f688746c8cdd/azure.png?h=250"  # noqa
    _description = "Run tasks using Azure Container Instances. Note this block is experimental. The interface may change without notice."  # noqa

    type: Literal["container-instance-job"] = Field(
        default="container-instance-job", description="The slug for this task type."
    )
    aci_credentials: AzureContainerInstanceCredentials
    resource_group_name: str = Field(
        default=...,
        title="Azure Resource Group Name",
        description=(
            "The name of the Azure Resource Group in which to run Prefect ACI tasks."
        ),
    )
    subscription_id: SecretStr = Field(
        default=...,
        title="Azure Subscription ID",
        description="The ID of the Azure subscription to create containers under.",
    )
    identities: Optional[List[str]] = Field(
        title="Identities",
        default=None,
        description=(
            "A list of user-assigned identities to associate with the container group. "
            "The identities should be an ARM resource IDs in the form: "
            "'/subscriptions/{subscriptionId}/resourceGroups/{resourceGroupName}/providers/Microsoft.ManagedIdentity/userAssignedIdentities/{identityName}'."  # noqa
        ),
    )
    image: Optional[str] = Field(
        default_factory=get_prefect_image_name,
        description=(
            "The image to use for the Prefect container in the task. This value "
            "defaults to a Prefect base image matching your local versions."
        ),
    )
    entrypoint: Optional[str] = Field(
        default=DEFAULT_CONTAINER_ENTRYPOINT,
        description=(
            "The entrypoint of the container you wish you run. This value "
            "defaults to the entrypoint used by Prefect images and should only be "
            "changed when using a custom image that is not based on an official "
            "Prefect image. Any commands set on deployments will be passed "
            "to the entrypoint as parameters."
        ),
    )
    image_registry: Optional[prefect.infrastructure.docker.DockerRegistry] = None
    cpu: float = Field(
        title="CPU",
        default=ACI_DEFAULT_CPU,
        description=(
            "The number of virtual CPUs to assign to the task container. "
            f"If not provided, a default value of {ACI_DEFAULT_CPU} will be used."
        ),
    )
    gpu_count: Optional[int] = Field(
        title="GPU Count",
        default=None,
        description=(
            "The number of GPUs to assign to the task container. "
            "If not provided, no GPU will be used."
        ),
    )
    gpu_sku: Optional[str] = Field(
        title="GPU SKU",
        default=None,
        description=(
            "The Azure GPU SKU to use. See the ACI documentation for a list of "
            "GPU SKUs available in each Azure region."
        ),
    )
    memory: float = Field(
        default=ACI_DEFAULT_MEMORY,
        description=(
            "The amount of memory in gigabytes to provide to the ACI task. Valid "
            "amounts are specified in the Azure documentation. If not provided, a "
            f"default value of  {ACI_DEFAULT_MEMORY} will be used unless present "
            "on the task definition."
        ),
    )
    subnet_ids: Optional[List[str]] = Field(
        default=None,
        title="Subnet IDs",
        description="A list of Azure subnet IDs the container should be connected to.",
    )
    dns_servers: Optional[List[str]] = Field(
        default=None,
        title="DNS Servers",
        description="A list of custom DNS Servers the container should use.",
    )
    stream_output: Optional[bool] = Field(
        default=None,
        description=(
            "If `True`, logs will be streamed from the Prefect container to the local "
            "console."
        ),
    )
    env: Dict[str, Optional[str]] = Field(
        title="Environment Variables",
        default_factory=dict,
        description=(
            "Environment variables to provide to the task run. These variables are set "
            "on the Prefect container at task runtime. These will not be set on the "
            "task definition."
        ),
    )
    # Execution settings
    task_start_timeout_seconds: int = Field(
        default=240,
        description=(
            "The amount of time to watch for the start of the ACI container. "
            "before marking it as failed."
        ),
    )
    task_watch_poll_interval: float = Field(
        default=5.0,
        description=(
            "The number of seconds to wait between Azure API calls while monitoring "
            "the state of an Azure Container Instances task."
        ),
    )

    @sync_compatible
    async def run(
        self, task_status: Optional[TaskStatus] = None
    ) -> AzureContainerInstanceJobResult:
        """
        Runs the configured task using an ACI container.

        Args:
            task_status: An optional `TaskStatus` to update when the container starts.

        Returns:
            An `AzureContainerInstanceJobResult` with the container's exit code.
        """

        run_start_time = datetime.datetime.now(datetime.timezone.utc)

        container = self._configure_container()
        container_group = self._configure_container_group(container)
        created_container_group = None

        aci_client = self.aci_credentials.get_container_client(
            self.subscription_id.get_secret_value()
        )

        self.logger.info(
            f"{self._log_prefix}: Preparing to run command {' '.join(self.command)!r} "
            f"in container {container.name!r} ({self.image})..."
        )
        try:
            self.logger.info(f"{self._log_prefix}: Waiting for container creation...")
            # Create the container group and wait for it to start
            creation_status_poller = await run_sync_in_worker_thread(
                aci_client.container_groups.begin_create_or_update,
                self.resource_group_name,
                container.name,
                container_group,
            )
            created_container_group = await run_sync_in_worker_thread(
                self._wait_for_task_container_start, creation_status_poller
            )

            # If creation succeeded, group provisioning state should be 'Succeeded'
            # and the group should have a single container
            if self._provisioning_succeeded(created_container_group):
                self.logger.info(f"{self._log_prefix}: Running command...")
                if task_status:
                    task_status.started(value=created_container_group.name)
                status_code = await run_sync_in_worker_thread(
                    self._watch_task_and_get_exit_code,
                    aci_client,
                    created_container_group,
                    run_start_time,
                )
                self.logger.info(f"{self._log_prefix}: Completed command run.")
            else:
                raise RuntimeError(f"{self._log_prefix}: Container creation failed.")

        finally:
            if created_container_group:
                await self._wait_for_container_group_deletion(
                    aci_client, created_container_group
                )

        return AzureContainerInstanceJobResult(
            identifier=created_container_group.name, status_code=status_code
        )

    async def kill(
        self,
        container_group_name: str,
        grace_seconds: int = CONTAINER_GROUP_DELETION_TIMEOUT_SECONDS,
    ):
        """
        Kill a flow running in an ACI container group.

        Args:
            container_group_name: The container group name yielded by
                `AzureContainerInstanceJob.run`.
        """
        # ACI does not provide a way to specify grace period, but it gives
        # applications ~30 seconds to gracefully terminate before killing
        # a container group.
        if grace_seconds != CONTAINER_GROUP_DELETION_TIMEOUT_SECONDS:
            self.logger.warning(
                f"{self._log_prefix}: Kill grace period of {grace_seconds}s requested, "
                f"but ACI does not support grace period configuration."
            )

        aci_client = self.aci_credentials.get_container_client(
            self.subscription_id.get_secret_value()
        )

        # get the container group to check that it still exists
        try:
            container_group = aci_client.container_groups.get(
                resource_group_name=self.resource_group_name,
                container_group_name=container_group_name,
            )
        except ResourceNotFoundError as exc:
            # the container group no longer exists, so there's nothing to cancel
            raise InfrastructureNotFound(
                f"Cannot stop ACI job: container group {container_group_name} "
                "no longer exists."
            ) from exc

        # get the container state to check if the container has terminated
        container = self._get_container(container_group)
        container_state = container.instance_view.current_state.state

        # the container group needs to be deleted regardless of whether the container
        # already terminated
        await self._wait_for_container_group_deletion(aci_client, container_group)

        # if the container had already terminated, raise an exception to let the agent
        # know the flow was not cancelled
        if container_state == ContainerRunState.TERMINATED:
            raise InfrastructureNotAvailable(
                f"Cannot stop ACI job: container group {container_group.name} exists, "
                f"but container {container.name} has already terminated."
            )

    def preview(self) -> str:
        """
        Provides a summary of how the container will be created when `run` is called.

        Returns:
           A string containing the summary.
        """
        preview = {
            "container_name": "<generated when run>",
            "resource_group_name": self.resource_group_name,
            "memory": self.memory,
            "cpu": self.cpu,
            "gpu_count": self.gpu_count,
            "gpu_sku": self.gpu_sku,
            "env": self._get_environment(),
        }

        return json.dumps(preview)

    def _configure_container(self) -> Container:
        """
        Configures an Azure `Container` using data from the block's fields.

        Returns:
            An instance of `Container` ready to submit to Azure.
        """

        # setup container environment variables
        environment = [
            EnvironmentVariable(name=k, secure_value=v)
            if k in ENV_SECRETS
            else EnvironmentVariable(name=k, value=v)
            for (k, v) in self._get_environment().items()
        ]

        # all container names in a resource group must be unique
        container_name = str(uuid.uuid4())
        container_resource_requirements = self._configure_container_resources()

        # add the entrypoint if provided, because creating an ACI container with a
        # command overrides the container's built-in entrypoint.
        if self.entrypoint:
            self.command.insert(0, self.entrypoint)

        return Container(
            name=container_name,
            image=self.image,
            command=self.command,
            resources=container_resource_requirements,
            environment_variables=environment,
        )

    def _configure_container_resources(self) -> ResourceRequirements:
        """
        Configures the container's memory, CPU, and GPU resources.

        Returns:
            A `ResourceRequirements` instance initialized with data from this
            `AzureContainerInstanceJob` block.
        """

        gpu_resource = (
            GpuResource(count=self.gpu_count, sku=self.gpu_sku)
            if self.gpu_count and self.gpu_sku
            else None
        )
        container_resource_requests = ResourceRequests(
            memory_in_gb=self.memory, cpu=self.cpu, gpu=gpu_resource
        )

        return ResourceRequirements(requests=container_resource_requests)

    def _configure_container_group(self, container: Container) -> ContainerGroup:
        """
        Configures the container group needed to start a container on ACI.

        Args:
            container: An initialized instance of `Container`.

        Returns:
            An initialized `ContainerGroup` ready to submit to Azure.
        """

        # Load the resource group, so we can set the container group location
        # correctly.

        resource_group_client = self.aci_credentials.get_resource_client(
            self.subscription_id.get_secret_value()
        )

        resource_group = resource_group_client.resource_groups.get(
            self.resource_group_name
        )

        image_registry_credentials = (
            [
                ImageRegistryCredential(
                    server=self.image_registry.registry_url,
                    username=self.image_registry.username,
                    password=self.image_registry.password.get_secret_value(),
                )
            ]
            if self.image_registry
            else None
        )

        identity = (
            ContainerGroupIdentity(
                type="UserAssigned",
                # The Azure API only uses the dict keys and ignores values when
                # creating a container group. Using empty `UserAssignedIdentities`
                # instances as dict values satisfies Python type checkers.
                user_assigned_identities={
                    identity: UserAssignedIdentities() for identity in self.identities
                },
            )
            if self.identities
            else None
        )

        subnet_ids = (
            [ContainerGroupSubnetId(id=subnet_id) for subnet_id in self.subnet_ids]
            if self.subnet_ids
            else None
        )

        dns_config = (
            DnsConfiguration(name_servers=self.dns_servers)
            if self.dns_servers
            else None
        )

        return ContainerGroup(
            location=resource_group.location,
            identity=identity,
            containers=[container],
            os_type=OperatingSystemTypes.linux,
            restart_policy=ContainerGroupRestartPolicy.never,
            image_registry_credentials=image_registry_credentials,
            subnet_ids=subnet_ids,
            dns_config=dns_config,
        )

    def _wait_for_task_container_start(
        self, creation_status_poller: LROPoller[ContainerGroup]
    ) -> ContainerGroup:
        """
        Wait for the result of group and container creation.

        Args:
            creation_status_poller: Poller returned by the Azure SDK.

        Raises:
            RuntimeError: Raised if the timeout limit is exceeded before the
            container starts.

        Returns:
            A `ContainerGroup` representing the current status of the group being
            watched.
        """

        t0 = time.time()
        timeout = self.task_start_timeout_seconds

        while not creation_status_poller.done():
            elapsed_time = time.time() - t0

            if timeout and elapsed_time > timeout:
                raise RuntimeError(
                    (
                        f"Timed out after {elapsed_time}s while watching waiting for "
                        "container start."
                    )
                )
            time.sleep(self.task_watch_poll_interval)

        return creation_status_poller.result()

    def _watch_task_and_get_exit_code(
        self,
        client: ContainerInstanceManagementClient,
        container_group: ContainerGroup,
        run_start_time: datetime.datetime,
    ) -> int:
        """
        Waits until the container finishes running and obtains its exit code.

        Args:
            client: An initialized Azure `ContainerInstanceManagementClient`
            container_group: The `ContainerGroup` in which the container resides.

        Returns:
            An `int` representing the container's exit code.
        """

        status_code = -1
        running_container = self._get_container(container_group)
        current_state = running_container.instance_view.current_state.state

        # get any logs the container has already generated
        last_log_time = run_start_time
        if self.stream_output:
            last_log_time = self._get_and_stream_output(
                client, container_group, last_log_time
            )

        # set exit code if flow run already finished:
        if current_state == ContainerRunState.TERMINATED:
            status_code = running_container.instance_view.current_state.exit_code

        while current_state != ContainerRunState.TERMINATED:
            try:
                container_group = client.container_groups.get(
                    resource_group_name=self.resource_group_name,
                    container_group_name=container_group.name,
                )
            except ResourceNotFoundError:
                self.logger.exception(
                    f"{self._log_prefix}: Container group was deleted before flow run "
                    "completed, likely due to flow cancellation."
                )

                # since the flow was cancelled, exit early instead of raising an
                # exception
                return status_code

            container = self._get_container(container_group)
            current_state = container.instance_view.current_state.state

            if current_state == ContainerRunState.TERMINATED:
                status_code = container.instance_view.current_state.exit_code
                # break instead of waiting for next loop iteration because
                # trying to read logs from a terminated container raises an exception
                break

            if self.stream_output:
                last_log_time = self._get_and_stream_output(
                    client, container_group, last_log_time
                )

            time.sleep(self.task_watch_poll_interval)

        return status_code

    async def _wait_for_container_group_deletion(
        self,
        aci_client: ContainerInstanceManagementClient,
        container_group: ContainerGroup,
    ):
        self.logger.info(f"{self._log_prefix}: Deleting container...")

        deletion_status_poller = await run_sync_in_worker_thread(
            aci_client.container_groups.begin_delete,
            resource_group_name=self.resource_group_name,
            container_group_name=container_group.name,
        )

        t0 = time.time()
        timeout = CONTAINER_GROUP_DELETION_TIMEOUT_SECONDS

        while not deletion_status_poller.done():
            elapsed_time = time.time() - t0

            if timeout and elapsed_time > timeout:
                raise RuntimeError(
                    (
                        f"Timed out after {elapsed_time}s while waiting for deletion of"
                        f" container group {container_group.name}. To verify the group "
                        "has been deleted, check the Azure Portal or run "
                        f"az container show --name {container_group.name} --resource-group {self.resource_group_name}"  # noqa
                    )
                )
            await anyio.sleep(self.task_watch_poll_interval)

        self.logger.info(f"{self._log_prefix}: Container deleted.")

    def _get_container(self, container_group: ContainerGroup) -> Container:
        """
        Extracts the job container from a container group.
        """
        return container_group.containers[0]

    def _get_and_stream_output(
        self,
        client: ContainerInstanceManagementClient,
        container_group: ContainerGroup,
        last_log_time: datetime.datetime,
    ) -> datetime.datetime:
        """
        Fetches logs output from the job container and writes all entries after
        a given time to stderr.

        Args:
            client: An initialized `ContainerInstanceManagementClient`
            container_group: The container group that holds the job container.
            last_log_time: The timestamp of the last output line already streamed.

        Returns:
            The time of the most recent output line written by this call.
        """
        logs = self._get_logs(client, container_group)
        return self._stream_output(logs, last_log_time)

    def _get_logs(
        self,
        client: ContainerInstanceManagementClient,
        container_group: ContainerGroup,
        max_lines: int = 100,
    ) -> str:
        """
        Gets the most container logs up to a given maximum.

        Args:
            client: An initialized `ContainerInstanceManagementClient`
            container_group: The container group that holds the job container.
            max_lines: The number of log lines to pull. Defaults to 100.

        Returns:
            A string containing the requested log entries, one per line.
        """
        container = self._get_container(container_group)

        logs: Union[Logs, None] = None
        try:
            logs = client.containers.list_logs(
                resource_group_name=self.resource_group_name,
                container_group_name=container_group.name,
                container_name=container.name,
                tail=max_lines,
                timestamps=True,
            )
        except HttpResponseError:
            # Trying to get logs when the container is under heavy CPU load sometimes
            # results in an error, but we won't want to raise an exception and stop
            # monitoring the flow. Instead, log the error and carry on so we can try to
            # get all missed logs on the next check.
            self.logger.warning(
                f"{self._log_prefix}: Unable to retrieve logs from container "
                f"{container.name}. Trying again in {self.task_watch_poll_interval}s"
            )

        return logs.content if logs else ""

    def _stream_output(
        self, log_content: Union[str, None], last_log_time: datetime.datetime
    ) -> datetime.datetime:
        """
        Writes each entry from a string of log lines to stderr.

        Args:
            log_content: A string containing Azure container logs.
            last_log_time: The timestamp of the last output line already streamed.

        Returns:
            The time of the most recent output line written by this call.
        """
        if not log_content:
            # nothing to stream
            return last_log_time

        log_lines = log_content.split("\n")

        last_written_time = last_log_time

        for log_line in log_lines:
            # skip if the line is blank or whitespace
            if not log_line.strip():
                continue

            line_parts = log_line.split(" ")
            # timestamp should always be before first space in line
            line_timestamp = line_parts[0]
            line = " ".join(line_parts[1:])

            try:
                line_time = dateutil.parser.parse(line_timestamp)
                if line_time > last_written_time:
                    self._write_output_line(line)
                    last_written_time = line_time
            except dateutil.parser.ParserError as e:
                self.logger.debug(
                    (
                        f"{self._log_prefix}: Unable to parse timestamp from Azure "
                        "log line: %s"
                    ),
                    log_line,
                    exc_info=e,
                )

        return last_written_time

    def _get_environment(self):
        """
        Generates a dictionary of all environment variables to send to the
        ACI container.
        """
        return {**self._base_environment(), **self.env}

    @property
    def _log_prefix(self) -> str:
        """
        Internal property for generating a prefix for logs where `name` may be null
        """
        if self.name is not None:
            return f"AzureContainerInstanceJob {self.name!r}"
        else:
            return "AzureContainerInstanceJob"

    @staticmethod
    def _provisioning_succeeded(container_group: ContainerGroup) -> bool:
        """
        Determines whether ACI container group provisioning was successful.

        Args:
            container_group: a container group returned by the Azure SDK.

        Returns:
            True if provisioning was successful, False otherwise.
        """
        if not container_group:
            return False

        return (
            container_group.provisioning_state
            == ContainerGroupProvisioningState.SUCCEEDED
            and len(container_group.containers) == 1
        )

    @staticmethod
    def _write_output_line(line: str):
        """
        Writes a line of output to stderr.
        """
        print(line, file=sys.stderr)

Functions

kill(container_group_name, grace_seconds=CONTAINER_GROUP_DELETION_TIMEOUT_SECONDS) async

Kill a flow running in an ACI container group.

Parameters:

Name Type Description Default
container_group_name str

The container group name yielded by AzureContainerInstanceJob.run.

required
Source code in prefect_azure/container_instance.py
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
async def kill(
    self,
    container_group_name: str,
    grace_seconds: int = CONTAINER_GROUP_DELETION_TIMEOUT_SECONDS,
):
    """
    Kill a flow running in an ACI container group.

    Args:
        container_group_name: The container group name yielded by
            `AzureContainerInstanceJob.run`.
    """
    # ACI does not provide a way to specify grace period, but it gives
    # applications ~30 seconds to gracefully terminate before killing
    # a container group.
    if grace_seconds != CONTAINER_GROUP_DELETION_TIMEOUT_SECONDS:
        self.logger.warning(
            f"{self._log_prefix}: Kill grace period of {grace_seconds}s requested, "
            f"but ACI does not support grace period configuration."
        )

    aci_client = self.aci_credentials.get_container_client(
        self.subscription_id.get_secret_value()
    )

    # get the container group to check that it still exists
    try:
        container_group = aci_client.container_groups.get(
            resource_group_name=self.resource_group_name,
            container_group_name=container_group_name,
        )
    except ResourceNotFoundError as exc:
        # the container group no longer exists, so there's nothing to cancel
        raise InfrastructureNotFound(
            f"Cannot stop ACI job: container group {container_group_name} "
            "no longer exists."
        ) from exc

    # get the container state to check if the container has terminated
    container = self._get_container(container_group)
    container_state = container.instance_view.current_state.state

    # the container group needs to be deleted regardless of whether the container
    # already terminated
    await self._wait_for_container_group_deletion(aci_client, container_group)

    # if the container had already terminated, raise an exception to let the agent
    # know the flow was not cancelled
    if container_state == ContainerRunState.TERMINATED:
        raise InfrastructureNotAvailable(
            f"Cannot stop ACI job: container group {container_group.name} exists, "
            f"but container {container.name} has already terminated."
        )
preview()

Provides a summary of how the container will be created when run is called.

Returns:

Type Description
str

A string containing the summary.

Source code in prefect_azure/container_instance.py
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
def preview(self) -> str:
    """
    Provides a summary of how the container will be created when `run` is called.

    Returns:
       A string containing the summary.
    """
    preview = {
        "container_name": "<generated when run>",
        "resource_group_name": self.resource_group_name,
        "memory": self.memory,
        "cpu": self.cpu,
        "gpu_count": self.gpu_count,
        "gpu_sku": self.gpu_sku,
        "env": self._get_environment(),
    }

    return json.dumps(preview)
run(task_status=None) async

Runs the configured task using an ACI container.

Parameters:

Name Type Description Default
task_status Optional[TaskStatus]

An optional TaskStatus to update when the container starts.

None

Returns:

Type Description
AzureContainerInstanceJobResult

An AzureContainerInstanceJobResult with the container's exit code.

Source code in prefect_azure/container_instance.py
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
@sync_compatible
async def run(
    self, task_status: Optional[TaskStatus] = None
) -> AzureContainerInstanceJobResult:
    """
    Runs the configured task using an ACI container.

    Args:
        task_status: An optional `TaskStatus` to update when the container starts.

    Returns:
        An `AzureContainerInstanceJobResult` with the container's exit code.
    """

    run_start_time = datetime.datetime.now(datetime.timezone.utc)

    container = self._configure_container()
    container_group = self._configure_container_group(container)
    created_container_group = None

    aci_client = self.aci_credentials.get_container_client(
        self.subscription_id.get_secret_value()
    )

    self.logger.info(
        f"{self._log_prefix}: Preparing to run command {' '.join(self.command)!r} "
        f"in container {container.name!r} ({self.image})..."
    )
    try:
        self.logger.info(f"{self._log_prefix}: Waiting for container creation...")
        # Create the container group and wait for it to start
        creation_status_poller = await run_sync_in_worker_thread(
            aci_client.container_groups.begin_create_or_update,
            self.resource_group_name,
            container.name,
            container_group,
        )
        created_container_group = await run_sync_in_worker_thread(
            self._wait_for_task_container_start, creation_status_poller
        )

        # If creation succeeded, group provisioning state should be 'Succeeded'
        # and the group should have a single container
        if self._provisioning_succeeded(created_container_group):
            self.logger.info(f"{self._log_prefix}: Running command...")
            if task_status:
                task_status.started(value=created_container_group.name)
            status_code = await run_sync_in_worker_thread(
                self._watch_task_and_get_exit_code,
                aci_client,
                created_container_group,
                run_start_time,
            )
            self.logger.info(f"{self._log_prefix}: Completed command run.")
        else:
            raise RuntimeError(f"{self._log_prefix}: Container creation failed.")

    finally:
        if created_container_group:
            await self._wait_for_container_group_deletion(
                aci_client, created_container_group
            )

    return AzureContainerInstanceJobResult(
        identifier=created_container_group.name, status_code=status_code
    )

AzureContainerInstanceJobResult

Bases: InfrastructureResult

The result of an AzureContainerInstanceJob run.

Source code in prefect_azure/container_instance.py
127
128
129
130
class AzureContainerInstanceJobResult(InfrastructureResult):
    """
    The result of an `AzureContainerInstanceJob` run.
    """

ContainerGroupProvisioningState

Bases: str, Enum

Terminal provisioning states for ACI container groups. Per the Azure docs, the states in this Enum are the only ones that can be relied on as dependencies.

Source code in prefect_azure/container_instance.py
108
109
110
111
112
113
114
115
class ContainerGroupProvisioningState(str, Enum):
    """
    Terminal provisioning states for ACI container groups. Per the Azure docs,
    the states in this Enum are the only ones that can be relied on as dependencies.
    """

    SUCCEEDED = "Succeeded"
    FAILED = "Failed"

ContainerRunState

Bases: str, Enum

Terminal run states for ACI containers.

Source code in prefect_azure/container_instance.py
118
119
120
121
122
123
124
class ContainerRunState(str, Enum):
    """
    Terminal run states for ACI containers.
    """

    RUNNING = "Running"
    TERMINATED = "Terminated"