dstack 0.18.42__py3-none-any.whl → 0.18.44__py3-none-any.whl

dstack/_internal/server/statics/static/media/entraID.d65d1f3e9486a8e56d24fc07b3230885.svg

@@ -0,0 +1,9 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<svg id="uuid-f8d4d392-7c12-4bd9-baff-66fbf7814b91" data-name="Layer 1" xmlns="http://www.w3.org/2000/svg" width="14" height="14" viewBox="0 0 18 18">
+  <path d="m3.802,14.032c.388.242,1.033.511,1.715.511.621,0,1.198-.18,1.676-.487,0,0,.001,0,.002-.001l1.805-1.128v4.073c-.286,0-.574-.078-.824-.234l-4.374-2.734Z" fill="#225086"/>
+  <path d="m7.853,1.507L.353,9.967c-.579.654-.428,1.642.323,2.111,0,0,2.776,1.735,3.126,1.954.388.242,1.033.511,1.715.511.621,0,1.198-.18,1.676-.487,0,0,.001,0,.002-.001l1.805-1.128-4.364-2.728,4.365-4.924V1s0,0,0,0c-.424,0-.847.169-1.147.507Z" fill="#6df"/>
+  <polygon points="4.636 10.199 4.688 10.231 9 12.927 9.001 12.927 9.001 12.927 9.001 5.276 9 5.275 4.636 10.199" fill="#cbf8ff"/>
+  <path d="m17.324,12.078c.751-.469.902-1.457.323-2.111l-4.921-5.551c-.397-.185-.842-.291-1.313-.291-.925,0-1.752.399-2.302,1.026l-.109.123h0s4.364,4.924,4.364,4.924h0s0,0,0,0l-4.365,2.728v4.073c.287,0,.573-.078.823-.234l7.5-4.688Z" fill="#074793"/>
+  <path d="m9.001,1v4.275s.109-.123.109-.123c.55-.627,1.377-1.026,2.302-1.026.472,0,.916.107,1.313.291l-2.579-2.909c-.299-.338-.723-.507-1.146-.507Z" fill="#0294e4"/>
+  <polygon points="13.365 10.199 13.365 10.199 13.365 10.199 9.001 5.276 9.001 12.926 13.365 10.199" fill="#96bcc2"/>
+</svg>

dstack/_internal/server/testing/common.py

@@ -16,7 +16,7 @@ from dstack._internal.core.models.configurations import (
 )
 from dstack._internal.core.models.envs import Env
 from dstack._internal.core.models.fleets import FleetConfiguration, FleetSpec, FleetStatus
-from dstack._internal.core.models.gateways import GatewayStatus
+from dstack._internal.core.models.gateways import GatewayComputeConfiguration, GatewayStatus
 from dstack._internal.core.models.instances import (
     Disk,
     Gpu,
@@ -69,6 +69,7 @@ from dstack._internal.server.models import (
     InstanceModel,
     JobMetricsPoint,
     JobModel,
+    JobPrometheusMetrics,
     PlacementGroupModel,
     PoolModel,
     ProjectModel,
@@ -420,6 +421,24 @@ async def create_gateway_compute(
     return gateway_compute
 
 
+def get_gateway_compute_configuration(
+    project_name: str = "test-project",
+    instance_name: str = "test-instance",
+    backend: BackendType = BackendType.AWS,
+    region: str = "us",
+    public_ip: bool = True,
+) -> GatewayComputeConfiguration:
+    return GatewayComputeConfiguration(
+        project_name=project_name,
+        instance_name=instance_name,
+        backend=backend,
+        region=region,
+        public_ip=public_ip,
+        ssh_key_pub="",
+        certificate=None,
+    )
+
+
 async def create_pool(
     session: AsyncSession,
     project: ProjectModel,
@@ -533,13 +552,7 @@ async def create_instance(
         requirements = Requirements(resources=ResourcesSpec(cpu=1))
 
     if instance_configuration is None:
-        instance_configuration = InstanceConfiguration(
-            project_name="test_proj",
-            instance_name="test_instance_name",
-            instance_id="test instance id",
-            ssh_keys=[],
-            user="test_user",
-        )
+        instance_configuration = get_instance_configuration()
 
     if volumes is None:
         volumes = []
@@ -581,6 +594,19 @@ async def create_instance(
     return im
 
 
+def get_instance_configuration(
+    project_name: str = "test-project",
+    instance_name: str = "test-instance",
+    user: str = "dstack-user",
+) -> InstanceConfiguration:
+    return InstanceConfiguration(
+        project_name=project_name,
+        instance_name=instance_name,
+        user=user,
+        ssh_keys=[],
+    )
+
+
 def get_instance_offer_with_availability(
     backend: BackendType = BackendType.AWS,
     region: str = "eu-west",
@@ -833,6 +859,22 @@ async def create_job_metrics_point(
     return jmp
 
 
+async def create_job_prometheus_metrics(
+    session: AsyncSession,
+    job: JobModel,
+    collected_at: datetime = datetime(2023, 1, 2, 3, 4, tzinfo=timezone.utc),
+    text: str = "# Prometheus metrics\n",
+):
+    metrics = JobPrometheusMetrics(
+        job_id=job.id,
+        collected_at=collected_at,
+        text=text,
+    )
+    session.add(metrics)
+    await session.commit()
+    return metrics
+
+
 def get_private_key_string() -> str:
     return """
 -----BEGIN RSA PRIVATE KEY-----

dstack/api/_public/runs.py

@@ -25,6 +25,7 @@ from dstack._internal.core.models.profiles import (
     ProfileRetryPolicy,
     SpotPolicy,
     TerminationPolicy,
+    UtilizationPolicy,
 )
 from dstack._internal.core.models.repos.base import Repo
 from dstack._internal.core.models.resources import ResourcesSpec
@@ -215,7 +216,7 @@ class Run(ABC):
                         start_time=next_start_time,
                         end_time=None,
                         descending=False,
-                        limit=100,
+                        limit=1000,
                         diagnose=diagnose,
                     ),
                 )
@@ -485,6 +486,7 @@ class RunCollection:
         resources: Optional[ResourcesSpec] = None,
         spot_policy: Optional[SpotPolicy] = None,
         retry_policy: Optional[ProfileRetryPolicy] = None,
+        utilization_policy: Optional[UtilizationPolicy] = None,
         max_duration: Optional[Union[int, str]] = None,
         max_price: Optional[float] = None,
         working_dir: Optional[str] = None,
@@ -535,6 +537,7 @@ class RunCollection:
             spot_policy=spot_policy,
             retry=None,
             retry_policy=retry_policy,
+            utilization_policy=utilization_policy,
             max_duration=max_duration,
             stop_duration=stop_duration,
             max_price=max_price,

dstack/api/server/_fleets.py

@@ -104,6 +104,8 @@ def _get_fleet_spec_excludes(fleet_spec: FleetSpec) -> Optional[_ExcludeDict]:
         profile_excludes.add("availability_zones")
     if fleet_spec.configuration.blocks == 1:
         configuration_excludes["blocks"] = True
+    if fleet_spec.profile is not None and fleet_spec.profile.utilization_policy is None:
+        profile_excludes.add("utilization_policy")
 
     if ssh_hosts_excludes:
         ssh_config_excludes["hosts"] = {"__all__": ssh_hosts_excludes}

dstack/api/server/_runs.py

@@ -185,6 +185,10 @@ def _get_run_spec_excludes(run_spec: RunSpec) -> Optional[dict]:
         and configuration.inactivity_duration is None
     ):
         configuration_excludes["inactivity_duration"] = True
+    if configuration.utilization_policy is None:
+        configuration_excludes["utilization_policy"] = True
+    if profile is not None and profile.utilization_policy is None:
+        profile_excludes.add("utilization_policy")
 
     if configuration_excludes:
         spec_excludes["configuration"] = configuration_excludes

dstack/api/utils.py

@@ -2,6 +2,7 @@ from pathlib import Path
 from typing import Optional, Tuple
 
 import yaml
+from pydantic import ValidationError
 
 from dstack._internal.core.errors import ConfigurationError
 from dstack._internal.core.models.configurations import AnyRunConfiguration
@@ -96,6 +97,8 @@ def _load_profile_from_path(profiles_path: Path, profile_name: Optional[str]) ->
             config = ProfilesConfig.parse_obj(yaml.safe_load(f))
     except FileNotFoundError:
         return None
+    except ValidationError as e:
+        raise ConfigurationError(e)
 
     if profile_name is None:
         return config.default()

dstack/version.py

@@ -1,3 +1,3 @@
-__version__ = "0.18.42"
+__version__ = "0.18.44"
 __is_release__ = True
-base_image = "0.6"
+base_image = "0.7"

{dstack-0.18.42.dist-info → dstack-0.18.44.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: dstack
-Version: 0.18.42
+Version: 0.18.44
 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
@@ -37,7 +37,8 @@ Requires-Dist: websocket-client
 Requires-Dist: python-multipart>=0.0.16
 Requires-Dist: filelock
 Requires-Dist: psutil
-Requires-Dist: gpuhunt<0.1.0,>=0.0.19
+Requires-Dist: gpuhunt<0.2.0,>=0.1.0
+Requires-Dist: argcomplete>=3.5.0
 Provides-Extra: all
 Requires-Dist: fastapi; extra == "all"
 Requires-Dist: starlette>=0.26.0; extra == "all"
@@ -59,6 +60,7 @@ Requires-Dist: alembic-postgresql-enum; extra == "all"
 Requires-Dist: asyncpg; extra == "all"
 Requires-Dist: cachetools; extra == "all"
 Requires-Dist: python-json-logger>=3.1.0; extra == "all"
+Requires-Dist: prometheus-client; extra == "all"
 Requires-Dist: grpcio>=1.50; extra == "all"
 Requires-Dist: boto3; extra == "all"
 Requires-Dist: botocore; extra == "all"
@@ -99,6 +101,7 @@ Requires-Dist: alembic-postgresql-enum; extra == "aws"
 Requires-Dist: asyncpg; extra == "aws"
 Requires-Dist: cachetools; extra == "aws"
 Requires-Dist: python-json-logger>=3.1.0; extra == "aws"
+Requires-Dist: prometheus-client; extra == "aws"
 Requires-Dist: grpcio>=1.50; extra == "aws"
 Requires-Dist: boto3; extra == "aws"
 Requires-Dist: botocore; extra == "aws"
@@ -123,6 +126,7 @@ Requires-Dist: alembic-postgresql-enum; extra == "azure"
 Requires-Dist: asyncpg; extra == "azure"
 Requires-Dist: cachetools; extra == "azure"
 Requires-Dist: python-json-logger>=3.1.0; extra == "azure"
+Requires-Dist: prometheus-client; extra == "azure"
 Requires-Dist: grpcio>=1.50; extra == "azure"
 Requires-Dist: azure-identity>=1.12.0; extra == "azure"
 Requires-Dist: azure-mgmt-subscription>=3.1.1; extra == "azure"
@@ -151,6 +155,7 @@ Requires-Dist: alembic-postgresql-enum; extra == "datacrunch"
 Requires-Dist: asyncpg; extra == "datacrunch"
 Requires-Dist: cachetools; extra == "datacrunch"
 Requires-Dist: python-json-logger>=3.1.0; extra == "datacrunch"
+Requires-Dist: prometheus-client; extra == "datacrunch"
 Requires-Dist: grpcio>=1.50; extra == "datacrunch"
 Requires-Dist: datacrunch; extra == "datacrunch"
 Provides-Extra: gateway
@@ -182,6 +187,7 @@ Requires-Dist: alembic-postgresql-enum; extra == "gcp"
 Requires-Dist: asyncpg; extra == "gcp"
 Requires-Dist: cachetools; extra == "gcp"
 Requires-Dist: python-json-logger>=3.1.0; extra == "gcp"
+Requires-Dist: prometheus-client; extra == "gcp"
 Requires-Dist: grpcio>=1.50; extra == "gcp"
 Requires-Dist: google-auth>=2.3.0; extra == "gcp"
 Requires-Dist: google-cloud-storage>=2.0.0; extra == "gcp"
@@ -211,6 +217,7 @@ Requires-Dist: alembic-postgresql-enum; extra == "kubernetes"
 Requires-Dist: asyncpg; extra == "kubernetes"
 Requires-Dist: cachetools; extra == "kubernetes"
 Requires-Dist: python-json-logger>=3.1.0; extra == "kubernetes"
+Requires-Dist: prometheus-client; extra == "kubernetes"
 Requires-Dist: grpcio>=1.50; extra == "kubernetes"
 Requires-Dist: kubernetes; extra == "kubernetes"
 Provides-Extra: lambda
@@ -234,6 +241,7 @@ Requires-Dist: alembic-postgresql-enum; extra == "lambda"
 Requires-Dist: asyncpg; extra == "lambda"
 Requires-Dist: cachetools; extra == "lambda"
 Requires-Dist: python-json-logger>=3.1.0; extra == "lambda"
+Requires-Dist: prometheus-client; extra == "lambda"
 Requires-Dist: grpcio>=1.50; extra == "lambda"
 Requires-Dist: boto3; extra == "lambda"
 Requires-Dist: botocore; extra == "lambda"
@@ -258,6 +266,7 @@ Requires-Dist: alembic-postgresql-enum; extra == "oci"
 Requires-Dist: asyncpg; extra == "oci"
 Requires-Dist: cachetools; extra == "oci"
 Requires-Dist: python-json-logger>=3.1.0; extra == "oci"
+Requires-Dist: prometheus-client; extra == "oci"
 Requires-Dist: grpcio>=1.50; extra == "oci"
 Requires-Dist: oci; extra == "oci"
 Provides-Extra: server
@@ -281,6 +290,7 @@ Requires-Dist: alembic-postgresql-enum; extra == "server"
 Requires-Dist: asyncpg; extra == "server"
 Requires-Dist: cachetools; extra == "server"
 Requires-Dist: python-json-logger>=3.1.0; extra == "server"
+Requires-Dist: prometheus-client; extra == "server"
 Requires-Dist: grpcio>=1.50; extra == "server"
 
 <div style="text-align: center;">
@@ -307,9 +317,9 @@ 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)