dstack 0.18.41__tar.gz → 0.18.43__tar.gz

{dstack-0.18.41 → dstack-0.18.43}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: dstack
-Version: 0.18.41
+Version: 0.18.43
 Summary: dstack is an open-source orchestration engine for running AI workloads on any cloud or on-premises.
 Home-page: https://dstack.ai
 Author: Andrey Cheptsov
@@ -52,9 +52,9 @@ for AI workloads both in the cloud and on-prem, speeding up the development, tra
 
 ## Major news ✨
 
+- [2025/02] [dstack 0.18.41: GPU blocks, Proxy jump, inactivity duration, and more](https://github.com/dstackai/dstack/releases/tag/0.18.41)
 - [2025/01] [dstack 0.18.38: Intel Gaudi](https://github.com/dstackai/dstack/releases/tag/0.18.38)
 - [2025/01] [dstack 0.18.35: Vultr](https://github.com/dstackai/dstack/releases/tag/0.18.35)
-- [2024/12] [dstack 0.18.32: TPU v6e](https://github.com/dstackai/dstack/releases/tag/0.18.32)
 - [2024/12] [dstack 0.18.30: AWS Capacity Reservations and Capacity Blocks](https://github.com/dstackai/dstack/releases/tag/0.18.30)
 - [2024/10] [dstack 0.18.21: Instance volumes](https://github.com/dstackai/dstack/releases/tag/0.18.21)
 - [2024/10] [dstack 0.18.18: Hardware metrics monitoring](https://github.com/dstackai/dstack/releases/tag/0.18.18)

{dstack-0.18.41 → dstack-0.18.43}/README.md

@@ -25,9 +25,9 @@ for AI workloads both in the cloud and on-prem, speeding up the development, tra
 
 ## Major news ✨
 
+- [2025/02] [dstack 0.18.41: GPU blocks, Proxy jump, inactivity duration, and more](https://github.com/dstackai/dstack/releases/tag/0.18.41)
 - [2025/01] [dstack 0.18.38: Intel Gaudi](https://github.com/dstackai/dstack/releases/tag/0.18.38)
 - [2025/01] [dstack 0.18.35: Vultr](https://github.com/dstackai/dstack/releases/tag/0.18.35)
-- [2024/12] [dstack 0.18.32: TPU v6e](https://github.com/dstackai/dstack/releases/tag/0.18.32)
 - [2024/12] [dstack 0.18.30: AWS Capacity Reservations and Capacity Blocks](https://github.com/dstackai/dstack/releases/tag/0.18.30)
 - [2024/10] [dstack 0.18.21: Instance volumes](https://github.com/dstackai/dstack/releases/tag/0.18.21)
 - [2024/10] [dstack 0.18.18: Hardware metrics monitoring](https://github.com/dstackai/dstack/releases/tag/0.18.18)

{dstack-0.18.41 → dstack-0.18.43}/setup.py

@@ -48,7 +48,8 @@ BASE_DEPS = [
     "python-multipart>=0.0.16",
     "filelock",
     "psutil",
-    "gpuhunt>=0.0.19,<0.1.0",
+    "gpuhunt>=0.1.0,<0.2.0",
+    "argcomplete>=3.5.0",
 ]
 
 GATEWAY_AND_SERVER_COMMON_DEPS = [

{dstack-0.18.41 → dstack-0.18.43}/src/dstack/_internal/cli/commands/__init__.py

@@ -5,6 +5,7 @@ from typing import List, Optional
 
 from rich_argparse import RichHelpFormatter
 
+from dstack._internal.cli.services.completion import ProjectNameCompleter
 from dstack._internal.cli.utils.common import configure_logging
 from dstack.api import Client
 
@@ -61,7 +62,7 @@ class APIBaseCommand(BaseCommand):
             help="The name of the project. Defaults to [code]$DSTACK_PROJECT[/]",
             metavar="NAME",
             default=os.getenv("DSTACK_PROJECT"),
-        )
+        ).completer = ProjectNameCompleter()
 
     def _command(self, args: argparse.Namespace):
         configure_logging()

{dstack-0.18.41 → dstack-0.18.43}/src/dstack/_internal/cli/commands/apply.py

@@ -1,6 +1,8 @@
 import argparse
 from pathlib import Path
 
+from argcomplete import FilesCompleter
+
 from dstack._internal.cli.commands import APIBaseCommand
 from dstack._internal.cli.services.configurators import (
     get_apply_configurator_class,
@@ -42,7 +44,7 @@ class ApplyCommand(APIBaseCommand):
             metavar="FILE",
             help="The path to the configuration file. Defaults to [code]$PWD/.dstack.yml[/]",
             dest="configuration_file",
-        )
+        ).completer = FilesCompleter(allowednames=["*.yml", "*.yaml"])
         self._parser.add_argument(
             "-y",
             "--yes",
@@ -57,7 +59,7 @@ class ApplyCommand(APIBaseCommand):
         self._parser.add_argument(
             "-d",
             "--detach",
-            help="Exit immediately after sumbitting configuration",
+            help="Exit immediately after submitting configuration",
             action="store_true",
         )
         repo_group = self._parser.add_argument_group("Repo Options")

{dstack-0.18.41 → dstack-0.18.43}/src/dstack/_internal/cli/commands/attach.py

@@ -6,6 +6,11 @@ from typing import Optional
 
 from dstack._internal.cli.commands import APIBaseCommand
 from dstack._internal.cli.services.args import port_mapping
+from dstack._internal.cli.services.completion import RunNameCompleter
+from dstack._internal.cli.services.configurators.run import (
+    get_run_exit_code,
+    print_finished_message,
+)
 from dstack._internal.cli.utils.common import console
 from dstack._internal.core.consts import DSTACK_RUNNER_HTTP_PORT
 from dstack._internal.core.errors import CLIError
@@ -57,7 +62,7 @@ class AttachCommand(APIBaseCommand):
             type=int,
             default=0,
         )
-        self._parser.add_argument("run_name")
+        self._parser.add_argument("run_name").completer = RunNameCompleter()
 
     def _command(self, args: argparse.Namespace):
         super()._command(args)
@@ -99,6 +104,21 @@ class AttachCommand(APIBaseCommand):
                     pass
         finally:
             run.detach()
+        # TODO: Handle run resubmissions similar to dstack apply
+
+        # After reading the logs, the run may not be marked as finished immediately.
+        # Give the run some time to transition to a finished state before exiting.
+        for _ in range(30):
+            run.refresh()
+            if run.status.is_finished():
+                print_finished_message(run)
+                exit(get_run_exit_code(run))
+            time.sleep(1)
+        console.print(
+            "[error]Lost run connection. Timed out waiting for run final status."
+            " Check `dstack ps` to see if it's done or failed."
+        )
+        exit(1)
 
 
 _IGNORED_PORTS = [DSTACK_RUNNER_HTTP_PORT]

dstack-0.18.43/src/dstack/_internal/cli/commands/completion.py

@@ -0,0 +1,20 @@
+import argcomplete
+
+from dstack._internal.cli.commands import BaseCommand
+
+
+class CompletionCommand(BaseCommand):
+    NAME = "completion"
+    DESCRIPTION = "Generate shell completion scripts"
+
+    def _register(self):
+        super()._register()
+        self._parser.add_argument(
+            "shell",
+            help="The shell to generate the completion script for",
+            choices=["bash", "zsh"],
+        )
+
+    def _command(self, args):
+        super()._command(args)
+        print(argcomplete.shellcode(["dstack"], shell=args.shell))

{dstack-0.18.41 → dstack-0.18.43}/src/dstack/_internal/cli/commands/delete.py

@@ -1,6 +1,8 @@
 import argparse
 from pathlib import Path
 
+from argcomplete import FilesCompleter
+
 from dstack._internal.cli.commands import APIBaseCommand
 from dstack._internal.cli.services.configurators import (
     get_apply_configurator_class,
@@ -22,7 +24,7 @@ class DeleteCommand(APIBaseCommand):
             metavar="FILE",
             help="The path to the configuration file. Defaults to [code]$PWD/.dstack.yml[/]",
             dest="configuration_file",
-        )
+        ).completer = FilesCompleter(allowednames=["*.yml", "*.yaml"])
         self._parser.add_argument(
             "-y",
             "--yes",

{dstack-0.18.41 → dstack-0.18.43}/src/dstack/_internal/cli/commands/fleet.py

@@ -4,6 +4,7 @@ import time
 from rich.live import Live
 
 from dstack._internal.cli.commands import APIBaseCommand
+from dstack._internal.cli.services.completion import FleetNameCompleter
 from dstack._internal.cli.utils.common import (
     LIVE_TABLE_PROVISION_INTERVAL_SECS,
     LIVE_TABLE_REFRESH_RATE_PER_SEC,
@@ -47,7 +48,7 @@ class FleetCommand(APIBaseCommand):
         delete_parser.add_argument(
             "name",
             help="The name of the fleet",
-        )
+        ).completer = FleetNameCompleter()
         delete_parser.add_argument(
             "-i",
             "--instance",

{dstack-0.18.41 → dstack-0.18.43}/src/dstack/_internal/cli/commands/gateway.py

@@ -4,6 +4,7 @@ import time
 from rich.live import Live
 
 from dstack._internal.cli.commands import APIBaseCommand
+from dstack._internal.cli.services.completion import GatewayNameCompleter
 from dstack._internal.cli.utils.common import (
     LIVE_TABLE_PROVISION_INTERVAL_SECS,
     LIVE_TABLE_REFRESH_RATE_PER_SEC,
@@ -59,7 +60,9 @@ class GatewayCommand(APIBaseCommand):
             "delete", help="Delete a gateway", formatter_class=self._parser.formatter_class
         )
         delete_parser.set_defaults(subfunc=self._delete)
-        delete_parser.add_argument("name", help="The name of the gateway")
+        delete_parser.add_argument(
+            "name", help="The name of the gateway"
+        ).completer = GatewayNameCompleter()
         delete_parser.add_argument(
             "-y", "--yes", action="store_true", help="Don't ask for confirmation"
         )
@@ -68,7 +71,9 @@ class GatewayCommand(APIBaseCommand):
             "update", help="Update a gateway", formatter_class=self._parser.formatter_class
         )
         update_parser.set_defaults(subfunc=self._update)
-        update_parser.add_argument("name", help="The name of the gateway")
+        update_parser.add_argument(
+            "name", help="The name of the gateway"
+        ).completer = GatewayNameCompleter()
         update_parser.add_argument(
             "--set-default", action="store_true", help="Set it the default gateway for the project"
         )

{dstack-0.18.41 → dstack-0.18.43}/src/dstack/_internal/cli/commands/logs.py

@@ -3,6 +3,7 @@ import sys
 from pathlib import Path
 
 from dstack._internal.cli.commands import APIBaseCommand
+from dstack._internal.cli.services.completion import RunNameCompleter
 from dstack._internal.core.errors import CLIError
 from dstack._internal.utils.logging import get_logger
 
@@ -33,7 +34,7 @@ class LogsCommand(APIBaseCommand):
         )
         self._parser.add_argument(
             "--replica",
-            help="The relica number. Defaults to 0.",
+            help="The replica number. Defaults to 0.",
             type=int,
             default=0,
         )
@@ -43,7 +44,7 @@ class LogsCommand(APIBaseCommand):
             type=int,
             default=0,
         )
-        self._parser.add_argument("run_name")
+        self._parser.add_argument("run_name").completer = RunNameCompleter(all=True)
 
     def _command(self, args: argparse.Namespace):
         super()._command(args)

{dstack-0.18.41 → dstack-0.18.43}/src/dstack/_internal/cli/commands/stats.py

@@ -7,6 +7,7 @@ from rich.live import Live
 from rich.table import Table
 
 from dstack._internal.cli.commands import APIBaseCommand
+from dstack._internal.cli.services.completion import RunNameCompleter
 from dstack._internal.cli.utils.common import (
     LIVE_TABLE_PROVISION_INTERVAL_SECS,
     LIVE_TABLE_REFRESH_RATE_PER_SEC,
@@ -25,7 +26,7 @@ class StatsCommand(APIBaseCommand):
 
     def _register(self):
         super()._register()
-        self._parser.add_argument("run_name")
+        self._parser.add_argument("run_name").completer = RunNameCompleter()
         self._parser.add_argument(
             "-w",
             "--watch",

{dstack-0.18.41 → dstack-0.18.43}/src/dstack/_internal/cli/commands/stop.py

@@ -1,6 +1,7 @@
 import argparse
 
 from dstack._internal.cli.commands import APIBaseCommand
+from dstack._internal.cli.services.completion import RunNameCompleter
 from dstack._internal.cli.utils.common import confirm_ask
 from dstack._internal.core.errors import CLIError
 
@@ -13,7 +14,7 @@ class StopCommand(APIBaseCommand):
         super()._register()
         self._parser.add_argument("-x", "--abort", action="store_true")
         self._parser.add_argument("-y", "--yes", action="store_true")
-        self._parser.add_argument("run_name")
+        self._parser.add_argument("run_name").completer = RunNameCompleter()
 
     def _command(self, args: argparse.Namespace):
         super()._command(args)

{dstack-0.18.41 → dstack-0.18.43}/src/dstack/_internal/cli/commands/volume.py

@@ -4,6 +4,7 @@ import time
 from rich.live import Live
 
 from dstack._internal.cli.commands import APIBaseCommand
+from dstack._internal.cli.services.completion import VolumeNameCompleter
 from dstack._internal.cli.utils.common import (
     LIVE_TABLE_PROVISION_INTERVAL_SECS,
     LIVE_TABLE_REFRESH_RATE_PER_SEC,
@@ -47,7 +48,7 @@ class VolumeCommand(APIBaseCommand):
         delete_parser.add_argument(
             "name",
             help="The name of the volume",
-        )
+        ).completer = VolumeNameCompleter()
         delete_parser.add_argument(
             "-y", "--yes", help="Don't ask for confirmation", action="store_true"
         )

{dstack-0.18.41 → dstack-0.18.43}/src/dstack/_internal/cli/main.py

@@ -1,10 +1,12 @@
 import argparse
 
+import argcomplete
 from rich.markup import escape
 from rich_argparse import RichHelpFormatter
 
 from dstack._internal.cli.commands.apply import ApplyCommand
 from dstack._internal.cli.commands.attach import AttachCommand
+from dstack._internal.cli.commands.completion import CompletionCommand
 from dstack._internal.cli.commands.config import ConfigCommand
 from dstack._internal.cli.commands.delete import DeleteCommand
 from dstack._internal.cli.commands.fleet import FleetCommand
@@ -72,9 +74,13 @@ def main():
     StatsCommand.register(subparsers)
     StopCommand.register(subparsers)
     VolumeCommand.register(subparsers)
+    CompletionCommand.register(subparsers)
+
+    argcomplete.autocomplete(parser, always_complete_options=False)
 
     args, unknown_args = parser.parse_known_args()
     args.unknown = unknown_args
+
     try:
         check_for_updates()
         get_ssh_client_info()

dstack-0.18.43/src/dstack/_internal/cli/services/completion.py

@@ -0,0 +1,86 @@
+import argparse
+import os
+from abc import ABC, abstractmethod
+from typing import Iterable, List, Optional
+
+import argcomplete
+from argcomplete.completers import BaseCompleter
+
+from dstack._internal.core.errors import ConfigurationError
+from dstack._internal.core.services.configs import ConfigManager
+from dstack.api import Client
+
+
+class BaseAPINameCompleter(BaseCompleter, ABC):
+    """
+    Base class for name completers that fetch resource names via the API.
+    """
+
+    def __init__(self):
+        super().__init__()
+
+    def get_api(self, parsed_args: argparse.Namespace) -> Optional[Client]:
+        argcomplete.debug(f"{self.__class__.__name__}: Retrieving API client")
+        project = getattr(parsed_args, "project", os.getenv("DSTACK_PROJECT"))
+        try:
+            return Client.from_config(project_name=project)
+        except ConfigurationError as e:
+            argcomplete.debug(f"{self.__class__.__name__}: Error initializing API client: {e}")
+            return None
+
+    def __call__(self, prefix: str, parsed_args: argparse.Namespace, **kwargs) -> List[str]:
+        api = self.get_api(parsed_args)
+        if api is None:
+            return []
+
+        argcomplete.debug(f"{self.__class__.__name__}: Fetching completions")
+        try:
+            resource_names = self.fetch_resource_names(api)
+            return [name for name in resource_names if name.startswith(prefix)]
+        except Exception as e:
+            argcomplete.debug(
+                f"{self.__class__.__name__}: Error fetching resource completions: {e}"
+            )
+            return []
+
+    @abstractmethod
+    def fetch_resource_names(self, api: Client) -> Iterable[str]:
+        """
+        Returns an iterable of resource names.
+        """
+        pass
+
+
+class RunNameCompleter(BaseAPINameCompleter):
+    def __init__(self, all: bool = False):
+        super().__init__()
+        self.all = all
+
+    def fetch_resource_names(self, api: Client) -> Iterable[str]:
+        return [r.name for r in api.runs.list(self.all)]
+
+
+class FleetNameCompleter(BaseAPINameCompleter):
+    def fetch_resource_names(self, api: Client) -> Iterable[str]:
+        return [r.name for r in api.client.fleets.list(api.project)]
+
+
+class VolumeNameCompleter(BaseAPINameCompleter):
+    def fetch_resource_names(self, api: Client) -> Iterable[str]:
+        return [r.name for r in api.client.volumes.list(api.project)]
+
+
+class GatewayNameCompleter(BaseAPINameCompleter):
+    def fetch_resource_names(self, api: Client) -> Iterable[str]:
+        return [r.name for r in api.client.gateways.list(api.project)]
+
+
+class ProjectNameCompleter(BaseCompleter):
+    """
+    Completer for local project names.
+    """
+
+    def __call__(self, prefix: str, parsed_args: argparse.Namespace, **kwargs) -> List[str]:
+        argcomplete.debug(f"{self.__class__.__name__}: Listing projects from ConfigManager")
+        projects = ConfigManager().list_projects()
+        return [p for p in projects if p.startswith(prefix)]

{dstack-0.18.41 → dstack-0.18.43}/src/dstack/_internal/cli/services/configurators/run.py

@@ -34,7 +34,6 @@ from dstack._internal.core.models.configurations import (
     BaseRunConfigurationWithPorts,
     DevEnvironmentConfiguration,
     PortMapping,
-    PythonVersion,
     RunConfigurationType,
     ServiceConfiguration,
     TaskConfiguration,
@@ -73,12 +72,6 @@ class BaseRunConfigurator(ApplyEnvVarsConfiguratorMixin, BaseApplyConfigurator):
     ):
         self.apply_args(conf, configurator_args, unknown_args)
         self.validate_gpu_vendor_and_image(conf)
-        if conf.python == PythonVersion.PY38:
-            logger.warning(
-                "Specifying [code]python: 3.8[/] in run configurations is deprecated"
-                " and will be forbidden in a future [code]dstack[/] release."
-                " Please upgrade your configuration to a newer Python version."
-            )
         if repo is None:
             repo = self.api.repos.load(Path.cwd())
         config_manager = ConfigManager()
@@ -238,8 +231,8 @@ class BaseRunConfigurator(ApplyEnvVarsConfiguratorMixin, BaseApplyConfigurator):
                         reattach = True
                         break
                     if run.status.is_finished():
-                        _print_finished_message(run)
-                        exit(_get_run_exit_code(run))
+                        print_finished_message(run)
+                        exit(get_run_exit_code(run))
                     time.sleep(1)
                 if not reattach:
                     console.print(
@@ -439,7 +432,7 @@ class RunWithPortsConfigurator(BaseRunConfigurator):
     ):
         super().apply_args(conf, args, unknown)
         if args.ports:
-            conf.ports = list(merge_ports(conf.ports, args.ports).values())
+            conf.ports = list(_merge_ports(conf.ports, args.ports).values())
 
 
 class TaskConfigurator(RunWithPortsConfigurator):
@@ -475,17 +468,17 @@ class ServiceConfigurator(BaseRunConfigurator):
         self.interpolate_run_args(conf.commands, unknown)
 
 
-def merge_ports(conf: List[PortMapping], args: List[PortMapping]) -> Dict[int, PortMapping]:
-    unique_ports_constraint([pm.container_port for pm in conf])
-    unique_ports_constraint([pm.container_port for pm in args])
+def _merge_ports(conf: List[PortMapping], args: List[PortMapping]) -> Dict[int, PortMapping]:
+    _unique_ports_constraint([pm.container_port for pm in conf])
+    _unique_ports_constraint([pm.container_port for pm in args])
     ports = {pm.container_port: pm for pm in conf}
     for pm in args:  # override conf
         ports[pm.container_port] = pm
-    unique_ports_constraint([pm.local_port for pm in ports.values() if pm.local_port is not None])
+    _unique_ports_constraint([pm.local_port for pm in ports.values() if pm.local_port is not None])
     return ports
 
 
-def unique_ports_constraint(ports: List[int]):
+def _unique_ports_constraint(ports: List[int]):
     used_ports = set()
     for i in ports:
         if i in used_ports:
@@ -514,7 +507,7 @@ def _print_service_urls(run: Run) -> None:
     console.print()
 
 
-def _print_finished_message(run: Run):
+def print_finished_message(run: Run):
     if run.status == RunStatus.DONE:
         console.print("[code]Done[/]")
         return
@@ -542,7 +535,7 @@ def _print_finished_message(run: Run):
     console.print(f"[error]{message}[/]")
 
 
-def _get_run_exit_code(run: Run) -> int:
+def get_run_exit_code(run: Run) -> int:
     if run.status == RunStatus.DONE:
         return 0
     return 1

{dstack-0.18.41 → dstack-0.18.43}/src/dstack/_internal/cli/utils/fleet.py

@@ -45,7 +45,11 @@ def get_fleets_table(
             status = instance.status.value
             total_blocks = instance.total_blocks
             busy_blocks = instance.busy_blocks
-            if total_blocks is not None and total_blocks > 1:
+            if (
+                instance.status in [InstanceStatus.IDLE, InstanceStatus.BUSY]
+                and total_blocks is not None
+                and total_blocks > 1
+            ):
                 status = f"{busy_blocks}/{total_blocks} {InstanceStatus.BUSY.value}"
             if (
                 instance.status in [InstanceStatus.IDLE, InstanceStatus.BUSY]

{dstack-0.18.41 → dstack-0.18.43}/src/dstack/_internal/cli/utils/volume.py

@@ -22,6 +22,8 @@ def get_volumes_table(
     if verbose:
         table.add_column("REGION")
     table.add_column("STATUS")
+    if verbose:
+        table.add_column("ATTACHED")
     table.add_column("CREATED")
     if verbose:
         table.add_column("ERROR")
@@ -37,11 +39,18 @@ def get_volumes_table(
                 and volume.provisioning_data.availability_zone is not None
             ):
                 region += f" ({volume.provisioning_data.availability_zone})"
+        attached = "-"
+        if volume.attachments is not None:
+            attached = ", ".join(
+                {va.instance.fleet_name for va in volume.attachments if va.instance.fleet_name}
+            )
+            attached = attached or "-"
         row = {
             "NAME": volume.name,
             "BACKEND": backend,
             "REGION": region,
             "STATUS": volume.status,
+            "ATTACHED": attached,
             "CREATED": format_date(volume.created_at),
             "ERROR": volume.status_message,
         }

{dstack-0.18.41 → dstack-0.18.43}/src/dstack/_internal/core/backends/aws/compute.py

@@ -11,8 +11,11 @@ from dstack._internal import settings
 from dstack._internal.core.backends.aws.config import AWSConfig
 from dstack._internal.core.backends.base.compute import (
     Compute,
+    generate_unique_gateway_instance_name,
+    generate_unique_instance_name,
+    generate_unique_volume_name,
     get_gateway_user_data,
-    get_instance_name,
+    get_job_instance_name,
     get_user_data,
     merge_tags,
 )
@@ -152,10 +155,12 @@ class AWSCompute(Compute):
         if zones is not None and len(zones) == 0:
             raise NoCapacityError("No eligible availability zones")
 
+        instance_name = generate_unique_instance_name(instance_config)
         tags = {
-            "Name": instance_config.instance_name,
+            "Name": instance_name,
             "owner": "dstack",
             "dstack_project": project_name,
+            "dstack_name": instance_config.instance_name,
             "dstack_user": instance_config.user,
         }
         tags = merge_tags(tags=tags, backend_tags=self.config.tags)
@@ -214,7 +219,7 @@ class AWSCompute(Compute):
                         disk_size=disk_size,
                         image_id=image_id,
                         instance_type=instance_offer.instance.name,
-                        iam_instance_profile_arn=None,
+                        iam_instance_profile=self.config.iam_instance_profile,
                         user_data=get_user_data(authorized_keys=instance_config.get_public_keys()),
                         tags=aws_resources.make_tags(tags),
                         security_group_id=aws_resources.create_security_group(
@@ -259,6 +264,9 @@ class AWSCompute(Compute):
                 )
             except botocore.exceptions.ClientError as e:
                 logger.warning("Got botocore.exceptions.ClientError: %s", e)
+                if e.response["Error"]["Code"] == "InvalidParameterValue":
+                    msg = e.response["Error"].get("Message", "")
+                    raise ComputeError(f"Invalid AWS request: {msg}")
                 continue
         raise NoCapacityError()
 
@@ -274,7 +282,7 @@ class AWSCompute(Compute):
         # TODO: run_job is the same for vm-based backends, refactor
         instance_config = InstanceConfiguration(
             project_name=run.project_name,
-            instance_name=get_instance_name(run, job),  # TODO: generate name
+            instance_name=get_job_instance_name(run, job),  # TODO: generate name
             ssh_keys=[
                 SSHKey(public=project_ssh_public_key.strip()),
             ],
@@ -342,10 +350,12 @@ class AWSCompute(Compute):
         ec2_resource = self.session.resource("ec2", region_name=configuration.region)
         ec2_client = self.session.client("ec2", region_name=configuration.region)
 
+        instance_name = generate_unique_gateway_instance_name(configuration)
         tags = {
-            "Name": configuration.instance_name,
+            "Name": instance_name,
             "owner": "dstack",
             "dstack_project": configuration.project_name,
+            "dstack_name": configuration.instance_name,
         }
         if settings.DSTACK_VERSION is not None:
             tags["dstack_version"] = settings.DSTACK_VERSION
@@ -373,7 +383,7 @@ class AWSCompute(Compute):
                 disk_size=10,
                 image_id=aws_resources.get_gateway_image_id(ec2_client),
                 instance_type="t2.micro",
-                iam_instance_profile_arn=None,
+                iam_instance_profile=None,
                 user_data=get_gateway_user_data(configuration.ssh_key_pub),
                 tags=tags,
                 security_group_id=security_group_id,
@@ -403,7 +413,7 @@ class AWSCompute(Compute):
 
         logger.debug("Creating ALB for gateway %s...", configuration.instance_name)
         response = elb_client.create_load_balancer(
-            Name=f"{configuration.instance_name}-lb",
+            Name=f"{instance_name}-lb",
             Subnets=subnets_ids,
             SecurityGroups=[security_group_id],
             Scheme="internet-facing" if configuration.public_ip else "internal",
@@ -418,7 +428,7 @@ class AWSCompute(Compute):
 
         logger.debug("Creating Target Group for gateway %s...", configuration.instance_name)
         response = elb_client.create_target_group(
-            Name=f"{configuration.instance_name}-tg",
+            Name=f"{instance_name}-tg",
             Protocol="HTTP",
             Port=80,
             VpcId=vpc_id,
@@ -535,11 +545,13 @@ class AWSCompute(Compute):
     def create_volume(self, volume: Volume) -> VolumeProvisioningData:
         ec2_client = self.session.client("ec2", region_name=volume.configuration.region)
 
+        volume_name = generate_unique_volume_name(volume)
         tags = {
-            "Name": volume.configuration.name,
+            "Name": volume_name,
             "owner": "dstack",
-            "dstack_user": volume.user,
             "dstack_project": volume.project_name,
+            "dstack_name": volume.name,
+            "dstack_user": volume.user,
         }
         tags = merge_tags(tags=tags, backend_tags=self.config.tags)
 
@@ -635,11 +647,12 @@ class AWSCompute(Compute):
         ec2_client = self.session.client("ec2", region_name=volume.configuration.region)
 
         logger.debug("Detaching EBS volume %s from instance %s", volume.volume_id, instance_id)
+        attachment_data = get_or_error(volume.get_attachment_data_for_instance(instance_id))
         try:
             ec2_client.detach_volume(
                 VolumeId=volume.volume_id,
                 InstanceId=instance_id,
-                Device=get_or_error(volume.attachment_data).device_name,
+                Device=attachment_data.device_name,
                 Force=force,
             )
         except botocore.exceptions.ClientError as e: