skypilot-nightly 1.0.0.dev2024053101__py3-none-any.whl → 1.0.0.dev2025022801__py3-none-any.whl

sky/clouds/do.py

@@ -0,0 +1,313 @@
+""" Digital Ocean Cloud. """
+
+import json
+import typing
+from typing import Dict, Iterator, List, Optional, Tuple, Union
+
+from sky import clouds
+from sky.adaptors import do
+from sky.clouds import service_catalog
+from sky.provision.do import utils as do_utils
+from sky.utils import registry
+from sky.utils import resources_utils
+
+if typing.TYPE_CHECKING:
+    from sky import resources as resources_lib
+
+_CREDENTIAL_FILE = 'config.yaml'
+
+
+@registry.CLOUD_REGISTRY.register(aliases=['digitalocean'])
+class DO(clouds.Cloud):
+    """Digital Ocean Cloud"""
+
+    _REPR = 'DO'
+    _CLOUD_UNSUPPORTED_FEATURES = {
+        clouds.CloudImplementationFeatures.CLONE_DISK_FROM_CLUSTER:
+            'Migrating '
+            f'disk is not supported in {_REPR}.',
+        clouds.CloudImplementationFeatures.SPOT_INSTANCE:
+            'Spot instances are '
+            f'not supported in {_REPR}.',
+        clouds.CloudImplementationFeatures.CUSTOM_DISK_TIER:
+            'Custom disk tiers'
+            f' is not supported in {_REPR}.',
+    }
+    # DO maximum node name length defined as <= 255
+    # https://docs.digitalocean.com/reference/api/api-reference/#operation/droplets_create
+    # 255 - 8 = 247 characters since
+    # our provisioner adds additional `-worker`.
+    _MAX_CLUSTER_NAME_LEN_LIMIT = 247
+    _regions: List[clouds.Region] = []
+
+    # Using the latest SkyPilot provisioner API to provision and check status.
+    PROVISIONER_VERSION = clouds.ProvisionerVersion.SKYPILOT
+    STATUS_VERSION = clouds.StatusVersion.SKYPILOT
+
+    @classmethod
+    def _unsupported_features_for_resources(
+        cls, resources: 'resources_lib.Resources'
+    ) -> Dict[clouds.CloudImplementationFeatures, str]:
+        """The features not supported based on the resources provided.
+
+        This method is used by check_features_are_supported() to check if the
+        cloud implementation supports all the requested features.
+
+        Returns:
+            A dict of {feature: reason} for the features not supported by the
+            cloud implementation.
+        """
+        del resources  # unused
+        return cls._CLOUD_UNSUPPORTED_FEATURES
+
+    @classmethod
+    def _max_cluster_name_length(cls) -> Optional[int]:
+        return cls._MAX_CLUSTER_NAME_LEN_LIMIT
+
+    @classmethod
+    def regions_with_offering(
+        cls,
+        instance_type: str,
+        accelerators: Optional[Dict[str, int]],
+        use_spot: bool,
+        region: Optional[str],
+        zone: Optional[str],
+    ) -> List[clouds.Region]:
+        assert zone is None, 'DO does not support zones.'
+        del accelerators, zone  # unused
+        if use_spot:
+            return []
+        regions = service_catalog.get_region_zones_for_instance_type(
+            instance_type, use_spot, 'DO')
+        if region is not None:
+            regions = [r for r in regions if r.name == region]
+        return regions
+
+    @classmethod
+    def get_vcpus_mem_from_instance_type(
+        cls,
+        instance_type: str,
+    ) -> Tuple[Optional[float], Optional[float]]:
+        return service_catalog.get_vcpus_mem_from_instance_type(instance_type,
+                                                                clouds='DO')
+
+    @classmethod
+    def zones_provision_loop(
+        cls,
+        *,
+        region: str,
+        num_nodes: int,
+        instance_type: str,
+        accelerators: Optional[Dict[str, int]] = None,
+        use_spot: bool = False,
+    ) -> Iterator[None]:
+        del num_nodes  # unused
+        regions = cls.regions_with_offering(instance_type,
+                                            accelerators,
+                                            use_spot,
+                                            region=region,
+                                            zone=None)
+        for r in regions:
+            assert r.zones is None, r
+            yield r.zones
+
+    def instance_type_to_hourly_cost(
+        self,
+        instance_type: str,
+        use_spot: bool,
+        region: Optional[str] = None,
+        zone: Optional[str] = None,
+    ) -> float:
+        return service_catalog.get_hourly_cost(
+            instance_type,
+            use_spot=use_spot,
+            region=region,
+            zone=zone,
+            clouds='DO',
+        )
+
+    def accelerators_to_hourly_cost(
+        self,
+        accelerators: Dict[str, int],
+        use_spot: bool,
+        region: Optional[str] = None,
+        zone: Optional[str] = None,
+    ) -> float:
+        """Returns the hourly cost of the accelerators, in dollars/hour."""
+        # the acc price is include in the instance price.
+        del accelerators, use_spot, region, zone  # unused
+        return 0.0
+
+    def get_egress_cost(self, num_gigabytes: float) -> float:
+        return 0.0
+
+    def __repr__(self):
+        return self._REPR
+
+    @classmethod
+    def get_default_instance_type(
+        cls,
+        cpus: Optional[str] = None,
+        memory: Optional[str] = None,
+        disk_tier: Optional[resources_utils.DiskTier] = None,
+    ) -> Optional[str]:
+        """Returns the default instance type for DO."""
+        return service_catalog.get_default_instance_type(cpus=cpus,
+                                                         memory=memory,
+                                                         disk_tier=disk_tier,
+                                                         clouds='DO')
+
+    @classmethod
+    def get_accelerators_from_instance_type(
+            cls, instance_type: str) -> Optional[Dict[str, Union[int, float]]]:
+        return service_catalog.get_accelerators_from_instance_type(
+            instance_type, clouds='DO')
+
+    @classmethod
+    def get_zone_shell_cmd(cls) -> Optional[str]:
+        return None
+
+    def make_deploy_resources_variables(
+            self,
+            resources: 'resources_lib.Resources',
+            cluster_name: resources_utils.ClusterName,
+            region: 'clouds.Region',
+            zones: Optional[List['clouds.Zone']],
+            num_nodes: int,
+            dryrun: bool = False) -> Dict[str, Optional[str]]:
+        del zones, dryrun, cluster_name
+
+        r = resources
+        acc_dict = self.get_accelerators_from_instance_type(r.instance_type)
+        if acc_dict is not None:
+            custom_resources = json.dumps(acc_dict, separators=(',', ':'))
+        else:
+            custom_resources = None
+        image_id = None
+        if (resources.image_id is not None and
+                resources.extract_docker_image() is None):
+            if None in resources.image_id:
+                image_id = resources.image_id[None]
+            else:
+                assert region.name in resources.image_id
+                image_id = resources.image_id[region.name]
+        return {
+            'instance_type': resources.instance_type,
+            'custom_resources': custom_resources,
+            'region': region.name,
+            **({
+                'image_id': image_id
+            } if image_id else {})
+        }
+
+    def _get_feasible_launchable_resources(
+        self, resources: 'resources_lib.Resources'
+    ) -> resources_utils.FeasibleResources:
+        """Returns a list of feasible resources for the given resources."""
+        if resources.use_spot:
+            # TODO: Add hints to all return values in this method to help
+            #  users understand why the resources are not launchable.
+            return resources_utils.FeasibleResources([], [], None)
+        if resources.instance_type is not None:
+            assert resources.is_launchable(), resources
+            resources = resources.copy(accelerators=None)
+            return resources_utils.FeasibleResources([resources], [], None)
+
+        def _make(instance_list):
+            resource_list = []
+            for instance_type in instance_list:
+                r = resources.copy(
+                    cloud=DO(),
+                    instance_type=instance_type,
+                    accelerators=None,
+                    cpus=None,
+                )
+                resource_list.append(r)
+            return resource_list
+
+        # Currently, handle a filter on accelerators only.
+        accelerators = resources.accelerators
+        if accelerators is None:
+            # Return a default instance type
+            default_instance_type = DO.get_default_instance_type(
+                cpus=resources.cpus,
+                memory=resources.memory,
+                disk_tier=resources.disk_tier)
+            if default_instance_type is None:
+                return resources_utils.FeasibleResources([], [], None)
+            else:
+                return resources_utils.FeasibleResources(
+                    _make([default_instance_type]), [], None)
+
+        assert len(accelerators) == 1, resources
+        acc, acc_count = list(accelerators.items())[0]
+        (instance_list, fuzzy_candidate_list) = (
+            service_catalog.get_instance_type_for_accelerator(
+                acc,
+                acc_count,
+                use_spot=resources.use_spot,
+                cpus=resources.cpus,
+                memory=resources.memory,
+                region=resources.region,
+                zone=resources.zone,
+                clouds='DO',
+            ))
+        if instance_list is None:
+            return resources_utils.FeasibleResources([], fuzzy_candidate_list,
+                                                     None)
+        return resources_utils.FeasibleResources(_make(instance_list),
+                                                 fuzzy_candidate_list, None)
+
+    @classmethod
+    def check_credentials(cls) -> Tuple[bool, Optional[str]]:
+        """Verify that the user has valid credentials for DO."""
+
+        try:
+            do.exceptions()
+        except ImportError as err:
+            return False, str(err)
+
+        try:
+            # attempt to make a CURL request for listing instances
+            do_utils.client().droplets.list()
+        except do.exceptions().HttpResponseError as err:
+            return False, str(err)
+        except do_utils.DigitalOceanError as err:
+            return False, str(err)
+
+        return True, None
+
+    def get_credential_file_mounts(self) -> Dict[str, str]:
+        try:
+            do_utils.client()
+            return {
+                f'~/.config/doctl/{_CREDENTIAL_FILE}': do_utils.CREDENTIALS_PATH
+            }
+        except do_utils.DigitalOceanError:
+            return {}
+
+    @classmethod
+    def get_current_user_identity(cls) -> Optional[List[str]]:
+        # NOTE: used for very advanced SkyPilot functionality
+        # Can implement later if desired
+        return None
+
+    @classmethod
+    def get_image_size(cls, image_id: str, region: Optional[str]) -> float:
+        del region
+        try:
+            response = do_utils.client().images.get(image_id=image_id)
+            return response['image']['size_gigabytes']
+        except do.exceptions().HttpResponseError as err:
+            raise do_utils.DigitalOceanError(
+                'HTTP error while retrieving size of '
+                f'image_id {response}: {err.error.message}') from err
+        except KeyError as err:
+            raise do_utils.DigitalOceanError(
+                f'No image_id `{image_id}` found') from err
+
+    def instance_type_exists(self, instance_type: str) -> bool:
+        return service_catalog.instance_type_exists(instance_type, 'DO')
+
+    def validate_region_zone(self, region: Optional[str], zone: Optional[str]):
+        return service_catalog.validate_region_zone(region, zone, clouds='DO')

sky/clouds/fluidstack.py

@@ -1,28 +1,28 @@
 """Fluidstack Cloud."""
-import json
 import os
 import typing
-from typing import Dict, Iterator, List, Optional, Tuple
+from typing import Dict, Iterator, List, Optional, Tuple, Union
 
 import requests
 
 from sky import clouds
-from sky import status_lib
 from sky.clouds import service_catalog
 from sky.provision.fluidstack import fluidstack_utils
+from sky.utils import registry
+from sky.utils import resources_utils
+from sky.utils import status_lib
 from sky.utils.resources_utils import DiskTier
 
 _CREDENTIAL_FILES = [
     # credential files for FluidStack,
-    fluidstack_utils.FLUIDSTACK_API_KEY_PATH,
-    fluidstack_utils.FLUIDSTACK_API_TOKEN_PATH,
+    fluidstack_utils.FLUIDSTACK_API_KEY_PATH
 ]
 if typing.TYPE_CHECKING:
     # Renaming to avoid shadowing variables.
     from sky import resources as resources_lib
 
 
-@clouds.CLOUD_REGISTRY.register
+@registry.CLOUD_REGISTRY.register
 class Fluidstack(clouds.Cloud):
     """FluidStack GPU Cloud."""
 
@@ -155,7 +155,7 @@ class Fluidstack(clouds.Cloud):
     def get_accelerators_from_instance_type(
         cls,
         instance_type: str,
-    ) -> Optional[Dict[str, int]]:
+    ) -> Optional[Dict[str, Union[int, float]]]:
         return service_catalog.get_accelerators_from_instance_type(
             instance_type, clouds='fluidstack')
 
@@ -174,9 +174,10 @@ class Fluidstack(clouds.Cloud):
     def make_deploy_resources_variables(
         self,
         resources: 'resources_lib.Resources',
-        cluster_name_on_cloud: str,
+        cluster_name: resources_utils.ClusterName,
         region: clouds.Region,
         zones: Optional[List[clouds.Zone]],
+        num_nodes: int,
         dryrun: bool = False,
     ) -> Dict[str, Optional[str]]:
 
@@ -184,24 +185,14 @@ class Fluidstack(clouds.Cloud):
 
         r = resources
         acc_dict = self.get_accelerators_from_instance_type(r.instance_type)
-        if acc_dict is not None:
-            custom_resources = json.dumps(acc_dict, separators=(',', ':'))
-        else:
-            custom_resources = None
-        cuda_installation_commands = """
-        sudo wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-keyring_1.1-1_all.deb -O /usr/local/cuda-keyring_1.1-1_all.deb; 
-        sudo dpkg -i /usr/local/cuda-keyring_1.1-1_all.deb;
-        sudo apt-get update;
-        sudo apt-get -y install cuda-toolkit-12-3;
-        sudo apt-get install -y cuda-drivers;
-        sudo apt-get install -y python3-pip;
-        nvidia-smi || sudo reboot;"""
+        custom_resources = resources_utils.make_ray_custom_resources_str(
+            acc_dict)
+
         return {
             'instance_type': resources.instance_type,
             'custom_resources': custom_resources,
             'region': region.name,
-            'fluidstack_username': self.default_username(region.name),
-            'cuda_installation_commands': cuda_installation_commands,
+            'fluidstack_username': 'ubuntu',
         }
 
     def _get_feasible_launchable_resources(
@@ -210,7 +201,9 @@ class Fluidstack(clouds.Cloud):
             assert resources.is_launchable(), resources
             # Accelerators are part of the instance type in Fluidstack Cloud
             resources = resources.copy(accelerators=None)
-            return ([resources], [])
+            # TODO: Add hints to all return values in this method to help
+            #  users understand why the resources are not launchable.
+            return resources_utils.FeasibleResources([resources], [], None)
 
         def _make(instance_list):
             resource_list = []
@@ -238,9 +231,10 @@ class Fluidstack(clouds.Cloud):
                 memory=resources.memory,
                 disk_tier=resources.disk_tier)
             if default_instance_type is None:
-                return ([], [])
+                return resources_utils.FeasibleResources([], [], None)
             else:
-                return (_make([default_instance_type]), [])
+                return resources_utils.FeasibleResources(
+                    _make([default_instance_type]), [], None)
 
         assert len(accelerators) == 1, resources
         acc, acc_count = list(accelerators.items())[0]
@@ -255,8 +249,10 @@ class Fluidstack(clouds.Cloud):
             zone=resources.zone,
             clouds='fluidstack')
         if instance_list is None:
-            return ([], fuzzy_candidate_list)
-        return (_make(instance_list), fuzzy_candidate_list)
+            return resources_utils.FeasibleResources([], fuzzy_candidate_list,
+                                                     None)
+        return resources_utils.FeasibleResources(_make(instance_list),
+                                                 fuzzy_candidate_list, None)
 
     @classmethod
     def check_credentials(cls) -> Tuple[bool, Optional[str]]:
@@ -264,17 +260,26 @@ class Fluidstack(clouds.Cloud):
         try:
             assert os.path.exists(
                 os.path.expanduser(fluidstack_utils.FLUIDSTACK_API_KEY_PATH))
-            assert os.path.exists(
-                os.path.expanduser(fluidstack_utils.FLUIDSTACK_API_TOKEN_PATH))
+
+            with open(os.path.expanduser(
+                    fluidstack_utils.FLUIDSTACK_API_KEY_PATH),
+                      encoding='UTF-8') as f:
+                api_key = f.read().strip()
+                if not api_key.startswith('api_key'):
+                    return False, ('Invalid FluidStack API key format. '
+                                   'To configure credentials, go to:\n    '
+                                   '  https://dashboard.fluidstack.io \n    '
+                                   'to obtain an API key, '
+                                   'then add save the contents '
+                                   'to ~/.fluidstack/api_key \n')
         except AssertionError:
-            return False, (
-                'Failed to access FluidStack Cloud'
-                ' with credentials. '
-                'To configure credentials, go to:\n    '
-                '  https://console.fluidstack.io \n    '
-                'to obtain an API key and API Token, '
-                'then add save the contents '
-                'to ~/.fluidstack/api_key and ~/.fluidstack/api_token \n')
+            return False, ('Failed to access FluidStack Cloud'
+                           ' with credentials. '
+                           'To configure credentials, go to:\n    '
+                           '  https://dashboard.fluidstack.io \n    '
+                           'to obtain an API key, '
+                           'then add save the contents '
+                           'to ~/.fluidstack/api_key \n')
         except requests.exceptions.ConnectionError:
             return False, ('Failed to verify FluidStack Cloud credentials. '
                            'Check your network connection '
@@ -285,8 +290,8 @@ class Fluidstack(clouds.Cloud):
         return {filename: filename for filename in _CREDENTIAL_FILES}
 
     @classmethod
-    def get_current_user_identity(cls) -> Optional[List[str]]:
-        # TODO(mjibril): Implement get_current_user_identity for Fluidstack
+    def get_user_identities(cls) -> Optional[List[List[str]]]:
+        # TODO(mjibril): Implement get_active_user_identity for Fluidstack
         return None
 
     def instance_type_exists(self, instance_type: str) -> bool:
@@ -297,21 +302,6 @@ class Fluidstack(clouds.Cloud):
                                                     zone,
                                                     clouds='fluidstack')
 
-    @classmethod
-    def default_username(cls, region: str) -> str:
-        return {
-            'norway_2_eu': 'ubuntu',
-            'calgary_1_canada': 'ubuntu',
-            'norway_3_eu': 'ubuntu',
-            'norway_4_eu': 'ubuntu',
-            'india_2': 'root',
-            'nevada_1_usa': 'fsuser',
-            'generic_1_canada': 'ubuntu',
-            'iceland_1_eu': 'ubuntu',
-            'new_york_1_usa': 'fsuser',
-            'illinois_1_usa': 'fsuser'
-        }.get(region, 'ubuntu')
-
     @classmethod
     def query_status(
         cls,