skypilot-nightly 1.0.0.dev20250915__py3-none-any.whl → 1.0.0.dev20250918__py3-none-any.whl

sky/skylet/services.py

@@ -10,7 +10,11 @@ from sky.schemas.generated import autostopv1_pb2
 from sky.schemas.generated import autostopv1_pb2_grpc
 from sky.schemas.generated import jobsv1_pb2
 from sky.schemas.generated import jobsv1_pb2_grpc
+from sky.schemas.generated import servev1_pb2
+from sky.schemas.generated import servev1_pb2_grpc
+from sky.serve import serve_rpc_utils
 from sky.serve import serve_state
+from sky.serve import serve_utils
 from sky.skylet import autostop_lib
 from sky.skylet import constants
 from sky.skylet import job_lib
@@ -52,6 +56,100 @@ class AutostopServiceImpl(autostopv1_pb2_grpc.AutostopServiceServicer):
             context.abort(grpc.StatusCode.INTERNAL, str(e))
 
 
+class ServeServiceImpl(servev1_pb2_grpc.ServeServiceServicer):
+    """Implementation of the ServeService gRPC service."""
+
+    # NOTE (kyuds): this grpc service will run cluster-side,
+    # thus guaranteeing that SERVE_VERSION is above 5.
+    # Therefore, we removed some SERVE_VERSION checks
+    # present in the original codegen.
+
+    def GetServiceStatus(  # type: ignore[return]
+            self, request: servev1_pb2.GetServiceStatusRequest,
+            context: grpc.ServicerContext
+    ) -> servev1_pb2.GetServiceStatusResponse:
+        """Gets serve status."""
+        try:
+            service_names, pool = (
+                serve_rpc_utils.GetServiceStatusRequestConverter.from_proto(request))  # pylint: disable=line-too-long
+            statuses = serve_utils.get_service_status_pickled(
+                service_names, pool)
+            return serve_rpc_utils.GetServiceStatusResponseConverter.to_proto(
+                statuses)
+        except Exception as e:  # pylint: disable=broad-except
+            context.abort(grpc.StatusCode.INTERNAL, str(e))
+
+    def AddVersion(  # type: ignore[return]
+            self, request: servev1_pb2.AddVersionRequest,
+            context: grpc.ServicerContext) -> servev1_pb2.AddVersionResponse:
+        """Adds serve version"""
+        try:
+            service_name = request.service_name
+            version = serve_state.add_version(service_name)
+            return servev1_pb2.AddVersionResponse(version=version)
+        except Exception as e:  # pylint: disable=broad-except
+            context.abort(grpc.StatusCode.INTERNAL, str(e))
+
+    def TerminateServices(  # type: ignore[return]
+            self, request: servev1_pb2.TerminateServicesRequest,
+            context: grpc.ServicerContext
+    ) -> servev1_pb2.TerminateServicesResponse:
+        """Terminates serve"""
+        try:
+            service_names, purge, pool = (
+                serve_rpc_utils.TerminateServicesRequestConverter.from_proto(request))  # pylint: disable=line-too-long
+            message = serve_utils.terminate_services(service_names, purge, pool)
+            return servev1_pb2.TerminateServicesResponse(message=message)
+        except Exception as e:  # pylint: disable=broad-except
+            context.abort(grpc.StatusCode.INTERNAL, str(e))
+
+    def TerminateReplica(  # type: ignore[return]
+            self, request: servev1_pb2.TerminateReplicaRequest,
+            context: grpc.ServicerContext
+    ) -> servev1_pb2.TerminateReplicaResponse:
+        """Terminate replica"""
+        try:
+            service_name = request.service_name
+            replica_id = request.replica_id
+            purge = request.purge
+            message = serve_utils.terminate_replica(service_name, replica_id,
+                                                    purge)
+            return servev1_pb2.TerminateReplicaResponse(message=message)
+        except Exception as e:  # pylint: disable=broad-except
+            context.abort(grpc.StatusCode.INTERNAL, str(e))
+
+    def WaitServiceRegistration(  # type: ignore[return]
+        self, request: servev1_pb2.WaitServiceRegistrationRequest,
+        context: grpc.ServicerContext
+    ) -> servev1_pb2.WaitServiceRegistrationResponse:
+        """Wait for service to be registered"""
+        try:
+            service_name = request.service_name
+            job_id = request.job_id
+            pool = request.pool
+            encoded = serve_utils.wait_service_registration(
+                service_name, job_id, pool)
+            lb_port = serve_utils.load_service_initialization_result(encoded)
+            return servev1_pb2.WaitServiceRegistrationResponse(lb_port=lb_port)
+        except Exception as e:  # pylint: disable=broad-except
+            context.abort(grpc.StatusCode.INTERNAL, str(e))
+
+    def UpdateService(  # type: ignore[return]
+            self, request: servev1_pb2.UpdateServiceRequest,
+            context: grpc.ServicerContext) -> servev1_pb2.UpdateServiceResponse:
+        """Update service"""
+        try:
+            service_name = request.service_name
+            version = request.version
+            mode = request.mode
+            pool = request.pool
+            serve_utils.update_service_encoded(service_name, version, mode,
+                                               pool)
+            return servev1_pb2.UpdateServiceResponse()
+        except Exception as e:  # pylint: disable=broad-except
+            context.abort(grpc.StatusCode.INTERNAL, str(e))
+
+
 class JobsServiceImpl(jobsv1_pb2_grpc.JobsServiceServicer):
     """Implementation of the JobsService gRPC service."""
 

sky/skylet/skylet.py

@@ -10,6 +10,7 @@ import sky
 from sky import sky_logging
 from sky.schemas.generated import autostopv1_pb2_grpc
 from sky.schemas.generated import jobsv1_pb2_grpc
+from sky.schemas.generated import servev1_pb2_grpc
 from sky.skylet import constants
 from sky.skylet import events
 from sky.skylet import services
@@ -50,9 +51,10 @@ def start_grpc_server(port: int = constants.SKYLET_GRPC_PORT) -> grpc.Server:
 
     autostopv1_pb2_grpc.add_AutostopServiceServicer_to_server(
         services.AutostopServiceImpl(), server)
-
     jobsv1_pb2_grpc.add_JobsServiceServicer_to_server(
         services.JobsServiceImpl(), server)
+    servev1_pb2_grpc.add_ServeServiceServicer_to_server(
+        services.ServeServiceImpl(), server)
 
     listen_addr = f'127.0.0.1:{port}'
     server.add_insecure_port(listen_addr)

sky/skypilot_config.py

@@ -415,10 +415,17 @@ def local_active_workspace_ctx(workspace: str) -> Iterator[None]:
 def get_active_workspace(force_user_workspace: bool = False) -> str:
     context_workspace = getattr(_active_workspace_context, 'workspace', None)
     if not force_user_workspace and context_workspace is not None:
-        logger.debug(f'Get context workspace: {context_workspace}')
+        logger.debug(f'Got context workspace: {context_workspace}')
         return context_workspace
-    return get_nested(keys=('active_workspace',),
-                      default_value=constants.SKYPILOT_DEFAULT_WORKSPACE)
+    active_workspace = get_nested(keys=('active_workspace',),
+                                  default_value=None)
+    if active_workspace is None:
+        logger.debug(f'No active workspace found, using default workspace: '
+                     f'{constants.SKYPILOT_DEFAULT_WORKSPACE}')
+        active_workspace = constants.SKYPILOT_DEFAULT_WORKSPACE
+    else:
+        logger.debug(f'Got active workspace: {active_workspace}')
+    return active_workspace
 
 
 def set_nested(keys: Tuple[str, ...], value: Any) -> Dict[str, Any]:

sky/templates/kubernetes-ray.yml.j2

@@ -823,7 +823,8 @@ available_node_types:
                 fi
                 $(prefix_cmd) cp -p "$FUSERMOUNT_PATH" "${FUSERMOUNT_PATH}-original"
                 $(prefix_cmd) ln -sf {{k8s_fusermount_shared_dir}}/fusermount-shim "$FUSERMOUNT_PATH"
-                FUSERMOUNT3_PATH=$(which fusermount3)
+                # "|| true" because fusermount3 is not always available
+                FUSERMOUNT3_PATH=$(which fusermount3) || true
                 if [ -z "$FUSERMOUNT3_PATH" ]; then
                     FUSERMOUNT3_PATH="${FUSERMOUNT_PATH}3"
                 fi
@@ -872,11 +873,14 @@ available_node_types:
 
                 touch /tmp/apt_ssh_setup_complete
                 echo "=== SSH setup completed ==="
-                ) > /tmp/${STEPS[0]}.log 2>&1 || {
-                  echo "Error: ${STEPS[0]} failed. Continuing anyway..." > /tmp/${STEPS[0]}.failed
+                ) > /tmp/${STEPS[0]}.log 2>&1
+                if [ "$?" -ne "0" ]; then
+                  {
+                  echo "Error: ${STEPS[0]} failed. Continuing anyway..." > /tmp/${STEPS[0]}.failed 2>&1
                   cat /tmp/${STEPS[0]}.log
                   exit 1
-                }
+                  }
+                fi
               ) &
 
               # STEP 2: Install conda, ray and skypilot (for dependencies); start
@@ -931,11 +935,14 @@ available_node_types:
                   set +e
                   {{ ray_worker_start_command }}
                 fi
-                ) > /tmp/${STEPS[1]}.log 2>&1 || {
-                  echo "Error: ${STEPS[1]} failed. Continuing anyway..." > /tmp/${STEPS[1]}.failed
+                ) > /tmp/${STEPS[1]}.log 2>&1
+                if [ "$?" -ne "0" ]; then
+                  {
+                  echo "Error: ${STEPS[1]} failed. Continuing anyway..." > /tmp/${STEPS[1]}.failed 2>&1
                   cat /tmp/${STEPS[1]}.log
                   exit 1
-                }
+                  }
+                fi
               ) &
 
 
@@ -953,11 +960,14 @@ available_node_types:
                   fi;
                 fi;
                 export -p > ~/container_env_var.sh && $(prefix_cmd) mv ~/container_env_var.sh /etc/profile.d/container_env_var.sh
-                ) > /tmp/${STEPS[2]}.log 2>&1 || {
-                  echo "Error: ${STEPS[2]} failed. Continuing anyway..." > /tmp/${STEPS[2]}.failed
+                ) > /tmp/${STEPS[2]}.log 2>&1
+                if [ "$?" -ne "0" ]; then
+                  {
+                  echo "Error: ${STEPS[2]} failed. Continuing anyway..." > /tmp/${STEPS[2]}.failed 2>&1
                   cat /tmp/${STEPS[2]}.log
                   exit 1
-                }
+                  }
+                fi
               ) &
 
               function mylsof { p=$(for pid in /proc/{0..9}*; do i=$(basename "$pid"); for file in "$pid"/fd/*; do link=$(readlink -e "$file"); if [ "$link" = "$1" ]; then echo "$i"; fi; done; done); echo "$p"; };
@@ -1324,7 +1334,7 @@ setup_commands:
     if [ -f /tmp/apt_ssh_setup_started ]; then
       echo "=== Logs for asynchronous SSH setup ===";
       [ -f /tmp/apt_ssh_setup_complete ] && cat /tmp/${STEPS[0]}.log ||
-      { tail -f -n +1 /tmp/${STEPS[0]}.log & TAIL_PID=$!; echo "Tail PID: $TAIL_PID"; until [ -f /tmp/apt_ssh_setup_complete ]; do sleep 0.5; done; kill $TAIL_PID || true; };
+      { tail -f -n +1 /tmp/${STEPS[0]}.log & TAIL_PID=$!; echo "Tail PID: $TAIL_PID"; until [ -f /tmp/apt_ssh_setup_complete ] || [ -f /tmp/${STEPS[0]}.failed ]; do sleep 0.5; done; kill $TAIL_PID || true; };
       [ -f /tmp/${STEPS[0]}.failed ] && { echo "Error: ${STEPS[0]} failed. Exiting."; exit 1; } || true;
     fi
 
@@ -1332,7 +1342,7 @@ setup_commands:
     if [ -f /tmp/skypilot_is_nimbus ]; then
       echo "=== Logs for asynchronous ray and skypilot installation ===";
       [ -f /tmp/ray_skypilot_installation_complete ] && cat /tmp/${STEPS[1]}.log ||
-      { tail -f -n +1 /tmp/${STEPS[1]}.log & TAIL_PID=$!; echo "Tail PID: $TAIL_PID"; until [ -f /tmp/ray_skypilot_installation_complete ]; do sleep 0.5; done; kill $TAIL_PID || true; };
+      { tail -f -n +1 /tmp/${STEPS[1]}.log & TAIL_PID=$!; echo "Tail PID: $TAIL_PID"; until [ -f /tmp/ray_skypilot_installation_complete ] || [ -f /tmp/${STEPS[1]}.failed ]; do sleep 0.5; done; kill $TAIL_PID || true; };
       [ -f /tmp/${STEPS[1]}.failed ] && { echo "Error: ${STEPS[1]} failed. Exiting."; exit 1; } || true;
     fi
     end_epoch=$(date +%s);

sky/templates/primeintellect-ray.yml.j2

@@ -0,0 +1,71 @@
+cluster_name: {{cluster_name_on_cloud}}
+
+# The maximum number of workers nodes to launch in addition to the head node.
+max_workers: {{num_nodes - 1}}
+upscaling_speed: {{num_nodes - 1}}
+idle_timeout_minutes: 60
+
+provider:
+  type: external
+  module: sky.provision.primeintellect
+  region: "{{region}}"
+  zones: "{{zones}}"
+
+auth:
+  ssh_user: skypilot:ssh_user
+  ssh_private_key: {{ssh_private_key}}
+
+available_node_types:
+  ray_head_default:
+    resources: {}
+    node_config:
+      InstanceType: {{instance_type}}
+      DiskSize: {{disk_size}}
+      ImageId: {{image_id}}
+      PublicKey: |-
+        skypilot:ssh_public_key_content
+
+head_node_type: ray_head_default
+
+# Format: `REMOTE_PATH : LOCAL_PATH`
+file_mounts: {
+  "{{sky_ray_yaml_remote_path}}": "{{sky_ray_yaml_local_path}}",
+  "{{sky_remote_path}}/{{sky_wheel_hash}}": "{{sky_local_path}}",
+{%- for remote_path, local_path in credentials.items() %}
+  "{{remote_path}}": "{{local_path}}",
+  "~/.ssh/sky-cluster-key": "{{ssh_private_key}}",
+{%- endfor %}
+}
+
+rsync_exclude: []
+
+initialization_commands: []
+
+# List of shell commands to run to set up nodes.
+# NOTE: these are very performance-sensitive. Each new item opens/closes an SSH
+# connection, which is expensive. Try your best to co-locate commands into fewer
+# items!
+#
+# Increment the following for catching performance bugs easier:
+#   current num items (num SSH connections): 1
+setup_commands:
+  # Disable unattended-upgrades and handle apt-get locks
+  # Install patch utility for Ray
+  # Install conda and Ray
+  # Set system limits for Ray performance (nofile and TasksMax)
+  - {%- for initial_setup_command in initial_setup_commands %}
+    {{ initial_setup_command }}
+    {%- endfor %}
+    sudo systemctl stop unattended-upgrades || true;
+    sudo systemctl disable unattended-upgrades || true;
+    sudo sed -i 's/Unattended-Upgrade "1"/Unattended-Upgrade "0"/g' /etc/apt/apt.conf.d/20auto-upgrades || true;
+    sudo kill -9 `sudo lsof /var/lib/dpkg/lock-frontend | awk '{print $2}' | tail -n 1` || true;
+    sudo pkill -9 apt-get;
+    sudo pkill -9 dpkg;
+    sudo dpkg --configure -a;
+    which patch > /dev/null || sudo apt install -y patch;
+    {{ conda_installation_commands }}
+    {{ ray_skypilot_installation_commands }}
+    sudo bash -c 'rm -rf /etc/security/limits.d; echo "* soft nofile 1048576" >> /etc/security/limits.conf; echo "* hard nofile 1048576" >> /etc/security/limits.conf';
+    sudo grep -e '^DefaultTasksMax' /etc/systemd/system.conf || (sudo bash -c 'echo "DefaultTasksMax=infinity" >> /etc/systemd/system.conf'); sudo systemctl set-property user-$(id -u $(whoami)).slice TasksMax=infinity; sudo systemctl daemon-reload;
+    {{ ssh_max_sessions_config }}

{skypilot_nightly-1.0.0.dev20250915.dist-info → skypilot_nightly-1.0.0.dev20250918.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: skypilot-nightly
-Version: 1.0.0.dev20250915
+Version: 1.0.0.dev20250918
 Summary: SkyPilot: Run AI on Any Infra — Unified, Faster, Cheaper.
 Author: SkyPilot Team
 License: Apache 2.0
@@ -121,6 +121,7 @@ Provides-Extra: fluidstack
 Provides-Extra: cudo
 Requires-Dist: cudo-compute>=0.1.10; extra == "cudo"
 Provides-Extra: paperspace
+Provides-Extra: primeintellect
 Provides-Extra: do
 Requires-Dist: pydo>=0.3.0; extra == "do"
 Requires-Dist: azure-core>=1.24.0; extra == "do"
@@ -151,49 +152,49 @@ Requires-Dist: grpcio>=1.63.0; extra == "server"
 Requires-Dist: protobuf<7.0.0,>=5.26.1; extra == "server"
 Requires-Dist: aiosqlite; extra == "server"
 Provides-Extra: all
-Requires-Dist: azure-identity>=1.19.0; extra == "all"
-Requires-Dist: python-dateutil; extra == "all"
-Requires-Dist: botocore>=1.29.10; extra == "all"
-Requires-Dist: awscli>=1.27.10; extra == "all"
-Requires-Dist: ibm-cos-sdk; extra == "all"
-Requires-Dist: anyio; extra == "all"
-Requires-Dist: msgraph-sdk; extra == "all"
-Requires-Dist: aiohttp; extra == "all"
-Requires-Dist: pyopenssl<24.3.0,>=23.2.0; extra == "all"
-Requires-Dist: pyjwt; extra == "all"
-Requires-Dist: passlib; extra == "all"
+Requires-Dist: casbin; extra == "all"
+Requires-Dist: kubernetes!=32.0.0,>=20.0.0; extra == "all"
+Requires-Dist: nebius>=0.2.47; extra == "all"
 Requires-Dist: ecsapi>=0.2.0; extra == "all"
+Requires-Dist: aiosqlite; extra == "all"
+Requires-Dist: protobuf<7.0.0,>=5.26.1; extra == "all"
+Requires-Dist: awscli>=1.27.10; extra == "all"
+Requires-Dist: docker; extra == "all"
+Requires-Dist: google-cloud-storage; extra == "all"
 Requires-Dist: azure-common; extra == "all"
-Requires-Dist: kubernetes!=32.0.0,>=20.0.0; extra == "all"
-Requires-Dist: azure-core>=1.31.0; extra == "all"
-Requires-Dist: azure-mgmt-compute>=33.0.0; extra == "all"
+Requires-Dist: ray[default]>=2.6.1; extra == "all"
+Requires-Dist: ibm-cloud-sdk-core; extra == "all"
+Requires-Dist: python-dateutil; extra == "all"
 Requires-Dist: azure-storage-blob>=12.23.1; extra == "all"
-Requires-Dist: boto3>=1.26.1; extra == "all"
-Requires-Dist: azure-core>=1.24.0; extra == "all"
-Requires-Dist: oci; extra == "all"
-Requires-Dist: docker; extra == "all"
+Requires-Dist: anyio; extra == "all"
+Requires-Dist: azure-identity>=1.19.0; extra == "all"
 Requires-Dist: sqlalchemy_adapter; extra == "all"
-Requires-Dist: google-cloud-storage; extra == "all"
-Requires-Dist: nebius>=0.2.47; extra == "all"
-Requires-Dist: azure-mgmt-network>=27.0.0; extra == "all"
-Requires-Dist: casbin; extra == "all"
+Requires-Dist: boto3>=1.26.1; extra == "all"
+Requires-Dist: pydo>=0.3.0; extra == "all"
+Requires-Dist: aiohttp; extra == "all"
 Requires-Dist: pyvmomi==8.0.1.0.2; extra == "all"
-Requires-Dist: grpcio>=1.63.0; extra == "all"
-Requires-Dist: ibm-platform-services>=0.48.0; extra == "all"
-Requires-Dist: vastai-sdk>=0.1.12; extra == "all"
-Requires-Dist: google-api-python-client>=2.69.0; extra == "all"
-Requires-Dist: ibm-vpc; extra == "all"
 Requires-Dist: colorama<0.4.5; extra == "all"
-Requires-Dist: pydo>=0.3.0; extra == "all"
+Requires-Dist: azure-core>=1.31.0; extra == "all"
 Requires-Dist: msrestazure; extra == "all"
-Requires-Dist: aiosqlite; extra == "all"
+Requires-Dist: cudo-compute>=0.1.10; extra == "all"
+Requires-Dist: oci; extra == "all"
+Requires-Dist: azure-core>=1.24.0; extra == "all"
+Requires-Dist: vastai-sdk>=0.1.12; extra == "all"
+Requires-Dist: ibm-vpc; extra == "all"
+Requires-Dist: botocore>=1.29.10; extra == "all"
+Requires-Dist: google-api-python-client>=2.69.0; extra == "all"
+Requires-Dist: ibm-platform-services>=0.48.0; extra == "all"
+Requires-Dist: msgraph-sdk; extra == "all"
+Requires-Dist: ibm-cos-sdk; extra == "all"
+Requires-Dist: azure-mgmt-network>=27.0.0; extra == "all"
+Requires-Dist: pyopenssl<24.3.0,>=23.2.0; extra == "all"
+Requires-Dist: azure-mgmt-compute>=33.0.0; extra == "all"
+Requires-Dist: grpcio>=1.63.0; extra == "all"
+Requires-Dist: pyjwt; extra == "all"
+Requires-Dist: websockets; extra == "all"
+Requires-Dist: passlib; extra == "all"
 Requires-Dist: runpod>=1.6.1; extra == "all"
 Requires-Dist: azure-cli>=2.65.0; extra == "all"
-Requires-Dist: protobuf<7.0.0,>=5.26.1; extra == "all"
-Requires-Dist: ibm-cloud-sdk-core; extra == "all"
-Requires-Dist: websockets; extra == "all"
-Requires-Dist: cudo-compute>=0.1.10; extra == "all"
-Requires-Dist: ray[default]>=2.6.1; extra == "all"
 Dynamic: author
 Dynamic: classifier
 Dynamic: description
@@ -291,12 +292,12 @@ SkyPilot supports your existing GPU, TPU, and CPU workloads, with no code change
 Install with pip:
 ```bash
 # Choose your clouds:
-pip install -U "skypilot[kubernetes,aws,gcp,azure,oci,nebius,lambda,runpod,fluidstack,paperspace,cudo,ibm,scp]"
+pip install -U "skypilot[kubernetes,aws,gcp,azure,oci,nebius,lambda,runpod,fluidstack,paperspace,cudo,ibm,scp,seeweb]"
 ```
 To get the latest features and fixes, use the nightly build or [install from source](https://docs.skypilot.co/en/latest/getting-started/installation.html):
 ```bash
 # Choose your clouds:
-pip install "skypilot-nightly[kubernetes,aws,gcp,azure,oci,nebius,lambda,runpod,fluidstack,paperspace,cudo,ibm,scp]"
+pip install "skypilot-nightly[kubernetes,aws,gcp,azure,oci,nebius,lambda,runpod,fluidstack,paperspace,cudo,ibm,scp,seeweb]"
 ```
 
 <p align="center">
@@ -305,7 +306,7 @@ pip install "skypilot-nightly[kubernetes,aws,gcp,azure,oci,nebius,lambda,runpod,
 
 Current supported infra: Kubernetes, AWS, GCP, Azure, OCI, Nebius, Lambda Cloud, RunPod, Fluidstack,
 Cudo, Digital Ocean, Paperspace, Cloudflare, Samsung, IBM, Vast.ai,
-VMware vSphere.
+VMware vSphere, Seeweb.
 <p align="center">
   <img alt="SkyPilot" src="https://raw.githubusercontent.com/skypilot-org/skypilot/master/docs/source/images/cloud-logos-light.png" width=85%>
 </p>