skypilot-nightly 1.0.0.dev20250106__tar.gz → 1.0.0.dev20250107__tar.gz

{skypilot_nightly-1.0.0.dev20250106/skypilot_nightly.egg-info → skypilot_nightly-1.0.0.dev20250107}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: skypilot-nightly
-Version: 1.0.0.dev20250106
+Version: 1.0.0.dev20250107
 Summary: SkyPilot: An intercloud broker for the clouds
 Author: SkyPilot Team
 License: Apache 2.0

{skypilot_nightly-1.0.0.dev20250106 → skypilot_nightly-1.0.0.dev20250107}/sky/__init__.py

@@ -5,7 +5,7 @@ from typing import Optional
 import urllib.request
 
 # Replaced with the current commit when building the wheels.
-_SKYPILOT_COMMIT_SHA = '38a822ac6b553df0e784e559715ee4269c21f780'
+_SKYPILOT_COMMIT_SHA = '6cf98a3cfbed39984b106f95ec7209058c73570d'
 
 
 def _get_git_commit():
@@ -35,7 +35,7 @@ def _get_git_commit():
 
 
 __commit__ = _get_git_commit()
-__version__ = '1.0.0.dev20250106'
+__version__ = '1.0.0.dev20250107'
 __root_dir__ = os.path.dirname(os.path.abspath(__file__))
 
 

{skypilot_nightly-1.0.0.dev20250106 → skypilot_nightly-1.0.0.dev20250107}/sky/backends/cloud_vm_ray_backend.py

@@ -4216,11 +4216,20 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
         attempts = 0
         while True:
             logger.debug(f'instance statuses attempt {attempts + 1}')
-            node_status_dict = provision_lib.query_instances(
-                repr(cloud),
-                cluster_name_on_cloud,
-                config['provider'],
-                non_terminated_only=False)
+            try:
+                node_status_dict = provision_lib.query_instances(
+                    repr(cloud),
+                    cluster_name_on_cloud,
+                    config['provider'],
+                    non_terminated_only=False)
+            except Exception as e:  # pylint: disable=broad-except
+                if purge:
+                    logger.warning(
+                        f'Failed to query instances. Skipping since purge is '
+                        f'set. Details: '
+                        f'{common_utils.format_exception(e, use_bracket=True)}')
+                    break
+                raise
 
             unexpected_node_state: Optional[Tuple[str, str]] = None
             for node_id, node_status in node_status_dict.items():
@@ -4239,8 +4248,13 @@ class CloudVmRayBackend(backends.Backend['CloudVmRayResourceHandle']):
                 time.sleep(_TEARDOWN_WAIT_BETWEEN_ATTEMPS_SECONDS)
             else:
                 (node_id, node_status) = unexpected_node_state
-                raise RuntimeError(f'Instance {node_id} in unexpected state '
-                                   f'{node_status}.')
+                if purge:
+                    logger.warning(f'Instance {node_id} in unexpected '
+                                   f'state {node_status}. Skipping since purge '
+                                   'is set.')
+                    break
+                raise RuntimeError(f'Instance {node_id} in unexpected '
+                                   f'state {node_status}.')
 
         global_user_state.remove_cluster(handle.cluster_name,
                                          terminate=terminate)

{skypilot_nightly-1.0.0.dev20250106 → skypilot_nightly-1.0.0.dev20250107}/sky/clouds/service_catalog/data_fetchers/fetch_gcp.py

@@ -47,10 +47,6 @@ TPU_RETRY_CNT = 3
 TPU_V4_ZONES = ['us-central2-b']
 # TPU v3 pods are available in us-east1-d, but hidden in the skus.
 # We assume the TPU prices are the same as us-central1.
-# TPU v6e's pricing info is not available on the SKUs. However, in
-# https://cloud.google.com/tpu/pricing, it listed the price for 4 regions:
-# us-east1, us-east5, europe-west4, and asia-northeast1. We hardcode them here
-# and filtered out the other regions (us-central{1,2}, us-south1).
 HIDDEN_TPU_DF = pd.read_csv(
     io.StringIO(
         textwrap.dedent("""\
@@ -62,49 +58,10 @@ HIDDEN_TPU_DF = pd.read_csv(
  ,tpu-v3-512,1,,,tpu-v3-512,512.0,153.6,us-east1,us-east1-d
  ,tpu-v3-1024,1,,,tpu-v3-1024,1024.0,307.2,us-east1,us-east1-d
  ,tpu-v3-2048,1,,,tpu-v3-2048,2048.0,614.4,us-east1,us-east1-d
- ,tpu-v6e-1,1,,,tpu-v6e-1,2.7,,us-east5,us-east5-b
- ,tpu-v6e-1,1,,,tpu-v6e-1,2.7,,us-east5,us-east5-c
- ,tpu-v6e-1,1,,,tpu-v6e-1,2.97,,europe-west4,europe-west4-a
- ,tpu-v6e-1,1,,,tpu-v6e-1,3.24,,asia-northeast1,asia-northeast1-b
- ,tpu-v6e-1,1,,,tpu-v6e-1,2.7,,us-east1,us-east1-d
- ,tpu-v6e-4,1,,,tpu-v6e-4,10.8,,us-east5,us-east5-b
- ,tpu-v6e-4,1,,,tpu-v6e-4,10.8,,us-east5,us-east5-c
- ,tpu-v6e-4,1,,,tpu-v6e-4,11.88,,europe-west4,europe-west4-a
- ,tpu-v6e-4,1,,,tpu-v6e-4,12.96,,asia-northeast1,asia-northeast1-b
- ,tpu-v6e-4,1,,,tpu-v6e-4,10.8,,us-east1,us-east1-d
- ,tpu-v6e-8,1,,,tpu-v6e-8,21.6,,us-east5,us-east5-b
- ,tpu-v6e-8,1,,,tpu-v6e-8,21.6,,us-east5,us-east5-c
- ,tpu-v6e-8,1,,,tpu-v6e-8,23.76,,europe-west4,europe-west4-a
- ,tpu-v6e-8,1,,,tpu-v6e-8,25.92,,asia-northeast1,asia-northeast1-b
- ,tpu-v6e-8,1,,,tpu-v6e-8,21.6,,us-east1,us-east1-d
- ,tpu-v6e-16,1,,,tpu-v6e-16,43.2,,us-east5,us-east5-b
- ,tpu-v6e-16,1,,,tpu-v6e-16,43.2,,us-east5,us-east5-c
- ,tpu-v6e-16,1,,,tpu-v6e-16,47.52,,europe-west4,europe-west4-a
- ,tpu-v6e-16,1,,,tpu-v6e-16,51.84,,asia-northeast1,asia-northeast1-b
- ,tpu-v6e-16,1,,,tpu-v6e-16,43.2,,us-east1,us-east1-d
- ,tpu-v6e-32,1,,,tpu-v6e-32,86.4,,us-east5,us-east5-b
- ,tpu-v6e-32,1,,,tpu-v6e-32,86.4,,us-east5,us-east5-c
- ,tpu-v6e-32,1,,,tpu-v6e-32,95.04,,europe-west4,europe-west4-a
- ,tpu-v6e-32,1,,,tpu-v6e-32,103.68,,asia-northeast1,asia-northeast1-b
- ,tpu-v6e-32,1,,,tpu-v6e-32,86.4,,us-east1,us-east1-d
- ,tpu-v6e-64,1,,,tpu-v6e-64,172.8,,us-east5,us-east5-b
- ,tpu-v6e-64,1,,,tpu-v6e-64,172.8,,us-east5,us-east5-c
- ,tpu-v6e-64,1,,,tpu-v6e-64,190.08,,europe-west4,europe-west4-a
- ,tpu-v6e-64,1,,,tpu-v6e-64,207.36,,asia-northeast1,asia-northeast1-b
- ,tpu-v6e-64,1,,,tpu-v6e-64,172.8,,us-east1,us-east1-d
- ,tpu-v6e-128,1,,,tpu-v6e-128,345.6,,us-east5,us-east5-b
- ,tpu-v6e-128,1,,,tpu-v6e-128,345.6,,us-east5,us-east5-c
- ,tpu-v6e-128,1,,,tpu-v6e-128,380.16,,europe-west4,europe-west4-a
- ,tpu-v6e-128,1,,,tpu-v6e-128,414.72,,asia-northeast1,asia-northeast1-b
- ,tpu-v6e-128,1,,,tpu-v6e-128,345.6,,us-east1,us-east1-d
- ,tpu-v6e-256,1,,,tpu-v6e-256,691.2,,us-east5,us-east5-b
- ,tpu-v6e-256,1,,,tpu-v6e-256,691.2,,us-east5,us-east5-c
- ,tpu-v6e-256,1,,,tpu-v6e-256,760.32,,europe-west4,europe-west4-a
- ,tpu-v6e-256,1,,,tpu-v6e-256,829.44,,asia-northeast1,asia-northeast1-b
- ,tpu-v6e-256,1,,,tpu-v6e-256,691.2,,us-east1,us-east1-d
  """)))
 
-TPU_V6E_MISSING_REGIONS = ['us-central1', 'us-central2', 'us-south1']
+# TPU V6e price for us-central2 is missing in the SKUs.
+TPU_V6E_MISSING_REGIONS = ['us-central2']
 
 # TPU V5 is not visible in specific zones. We hardcode the missing zones here.
 # NOTE(dev): Keep the zones and the df in sync.
@@ -670,6 +627,8 @@ def get_tpu_df(gce_skus: List[Dict[str, Any]],
                 return 'TpuV5p'
             assert tpu_version == 'v5litepod', tpu_version
             return 'TpuV5e'
+        if tpu_version.startswith('v6e'):
+            return 'TpuV6e'
         return f'Tpu-{tpu_version}'
 
     def get_tpu_price(row: pd.Series, spot: bool) -> Optional[float]:
@@ -684,10 +643,10 @@ def get_tpu_df(gce_skus: List[Dict[str, Any]],
         # whether the TPU is a single device or a pod.
         # For TPU-v4, the pricing is uniform, and thus the pricing API
         # only provides the price of TPU-v4 pods.
-        # The price shown for v5 TPU is per chip hour, so there is no 'Pod'
-        # keyword in the description.
+        # The price shown for v5 & v6e TPU is per chip hour, so there is
+        # no 'Pod' keyword in the description.
         is_pod = ((num_cores > 8 or tpu_version == 'v4') and
-                  not tpu_version.startswith('v5'))
+                  not tpu_version.startswith('v5') and tpu_version != 'v6e')
 
         for sku in gce_skus + tpu_skus:
             if tpu_region not in sku['serviceRegions']:
@@ -718,7 +677,9 @@ def get_tpu_df(gce_skus: List[Dict[str, Any]],
             # for v5e. Reference here:
             # https://cloud.google.com/tpu/docs/v5p#using-accelerator-type
             # https://cloud.google.com/tpu/docs/v5e#tpu-v5e-config
-            core_per_sku = (1 if tpu_version == 'v5litepod' else
+            # v6e is also per chip price. Reference here:
+            # https://cloud.google.com/tpu/docs/v6e#configurations
+            core_per_sku = (1 if tpu_version in ['v5litepod', 'v6e'] else
                             2 if tpu_version == 'v5p' else 8)
             tpu_core_price = tpu_device_price / core_per_sku
             tpu_price = num_cores * tpu_core_price
@@ -738,8 +699,6 @@ def get_tpu_df(gce_skus: List[Dict[str, Any]],
             spot_str = 'spot ' if spot else ''
             print(f'The {spot_str}price of {tpu_name} in {tpu_region} is '
                   'not found in SKUs or hidden TPU price DF.')
-        # TODO(tian): Hack. Should investigate how to retrieve the price
-        # for TPU-v6e.
         if (tpu_name.startswith('tpu-v6e') and
                 tpu_region in TPU_V6E_MISSING_REGIONS):
             if not spot:

{skypilot_nightly-1.0.0.dev20250106 → skypilot_nightly-1.0.0.dev20250107}/sky/clouds/utils/oci_utils.py

@@ -10,6 +10,8 @@ History:
    from ubuntu 20.04 to ubuntu 22.04, including:
    - GPU: skypilot:gpu-ubuntu-2004 -> skypilot:gpu-ubuntu-2204
    - CPU: skypilot:cpu-ubuntu-2004 -> skypilot:cpu-ubuntu-2204
+ - Hysun He (hysun.he@oracle.com) @ Jan.01, 2025: Support reuse existing
+   VCN for SkyServe.
 """
 import os
 
@@ -109,8 +111,15 @@ class OCIConfig:
             ('oci', region, 'compartment_ocid'), default_compartment_ocid)
         return compartment
 
+    @classmethod
+    def get_vcn_ocid(cls, region):
+        # Will reuse the regional VCN if specified.
+        vcn = skypilot_config.get_nested(('oci', region, 'vcn_ocid'), None)
+        return vcn
+
     @classmethod
     def get_vcn_subnet(cls, region):
+        # Will reuse the subnet if specified.
         vcn = skypilot_config.get_nested(('oci', region, 'vcn_subnet'), None)
         return vcn
 

{skypilot_nightly-1.0.0.dev20250106 → skypilot_nightly-1.0.0.dev20250107}/sky/provision/oci/query_utils.py

@@ -7,6 +7,8 @@ History:
    find_compartment: allow search subtree when find a compartment.
  - Hysun He (hysun.he@oracle.com) @ Nov.12, 2024: Add methods to
    Add/remove security rules: create_nsg_rules & remove_nsg
+ - Hysun He (hysun.he@oracle.com) @ Jan.01, 2025: Support reuse existing
+   VCN for SkyServe.
 """
 from datetime import datetime
 import functools
@@ -17,7 +19,6 @@ import traceback
 import typing
 from typing import List, Optional, Tuple
 
-from sky import exceptions
 from sky import sky_logging
 from sky.adaptors import common as adaptors_common
 from sky.adaptors import oci as oci_adaptor
@@ -496,26 +497,25 @@ class QueryHelper:
 
         compartment = cls.find_compartment(region)
 
-        list_vcns_resp = net_client.list_vcns(
-            compartment_id=compartment,
-            display_name=oci_utils.oci_config.VCN_NAME,
-            lifecycle_state='AVAILABLE',
-        )
-
-        if not list_vcns_resp:
-            raise exceptions.ResourcesUnavailableError(
-                'The VCN is not available')
+        vcn_id = oci_utils.oci_config.get_vcn_ocid(region)
+        if vcn_id is None:
+            list_vcns_resp = net_client.list_vcns(
+                compartment_id=compartment,
+                display_name=oci_utils.oci_config.VCN_NAME,
+                lifecycle_state='AVAILABLE',
+            )
 
-        # Get the primary vnic. The vnic might be an empty list for the
-        # corner case when the cluster was exited during provision.
-        if not list_vcns_resp.data:
-            return None
+            # Get the primary vnic. The vnic might be an empty list for the
+            # corner case when the cluster was exited during provision.
+            if not list_vcns_resp.data:
+                return None
 
-        vcn = list_vcns_resp.data[0]
+            vcn = list_vcns_resp.data[0]
+            vcn_id = vcn.id
 
         list_nsg_resp = net_client.list_network_security_groups(
             compartment_id=compartment,
-            vcn_id=vcn.id,
+            vcn_id=vcn_id,
             limit=1,
             display_name=nsg_name,
         )
@@ -532,7 +532,7 @@ class QueryHelper:
             create_network_security_group_details=oci_adaptor.oci.core.models.
             CreateNetworkSecurityGroupDetails(
                 compartment_id=compartment,
-                vcn_id=vcn.id,
+                vcn_id=vcn_id,
                 display_name=nsg_name,
             ))
         get_nsg_resp = net_client.get_network_security_group(

{skypilot_nightly-1.0.0.dev20250106 → skypilot_nightly-1.0.0.dev20250107}/sky/utils/schemas.py

@@ -886,6 +886,9 @@ def get_config_schema():
                     'image_tag_gpu': {
                         'type': 'string',
                     },
+                    'vcn_ocid': {
+                        'type': 'string',
+                    },
                     'vcn_subnet': {
                         'type': 'string',
                     },

{skypilot_nightly-1.0.0.dev20250106 → skypilot_nightly-1.0.0.dev20250107/skypilot_nightly.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: skypilot-nightly
-Version: 1.0.0.dev20250106
+Version: 1.0.0.dev20250107
 Summary: SkyPilot: An intercloud broker for the clouds
 Author: SkyPilot Team
 License: Apache 2.0