feat: Implment sync replica mutation in api layer

Author: seedspirit

src/ai/backend/manager/api/gql/model_deployment/model_deployment.py

@@ -37,6 +37,7 @@
 )
 from ai.backend.manager.api.gql.types import StrawberryGQLContext
 from ai.backend.manager.api.gql.user import User
+from ai.backend.manager.services.deployment.actions.sync_replicas import SyncReplicaAction
 
 from .model_revision import (
     CreateModelRevisionInput,
@@ -595,4 +596,9 @@ class SyncReplicaPayload:
 async def sync_replicas(
     input: SyncReplicaInput, info: Info[StrawberryGQLContext]
 ) -> SyncReplicaPayload:
+    deployment_processor = info.context.processors.deployment
+    assert deployment_processor is not None
+    await deployment_processor.sync_replicas.wait_for_complete(
+        SyncReplicaAction(deployment_id=UUID(input.model_deployment_id))
+    )
     return SyncReplicaPayload(success=True)

src/ai/backend/manager/services/deployment/actions/sync_replicas.py

@@ -0,0 +1,31 @@
+from dataclasses import dataclass
+from typing import Optional, override
+from uuid import UUID
+
+from ai.backend.manager.actions.action import BaseActionResult
+from ai.backend.manager.services.deployment.actions.base import DeploymentBaseAction
+
+
+@dataclass
+class SyncReplicaAction(DeploymentBaseAction):
+    """Action to sync replicas for an existing deployment."""
+
+    deployment_id: UUID
+
+    @override
+    def entity_id(self) -> Optional[str]:
+        return str(self.deployment_id)
+
+    @override
+    @classmethod
+    def operation_type(cls) -> str:
+        return "sync_replicas"
+
+
+@dataclass
+class SyncReplicaActionResult(BaseActionResult):
+    success: bool
+
+    @override
+    def entity_id(self) -> Optional[str]:
+        return None

src/ai/backend/manager/services/deployment/processors.py

@@ -25,6 +25,10 @@
     DestroyDeploymentAction,
     DestroyDeploymentActionResult,
 )
+from ai.backend.manager.services.deployment.actions.sync_replicas import (
+    SyncReplicaAction,
+    SyncReplicaActionResult,
+)
 from ai.backend.manager.services.deployment.actions.update_auto_scaling_rule import (
     UpdateAutoScalingRuleAction,
     UpdateAutoScalingRuleActionResult,
@@ -52,6 +56,8 @@ async def create_access_token(
         self, action: CreateAccessTokenAction
     ) -> CreateAccessTokenActionResult: ...
 
+    async def sync_replicas(self, action: SyncReplicaAction) -> SyncReplicaActionResult: ...
+
 
 class DeploymentProcessors(AbstractProcessorPackage):
     """Processors for deployment operations."""
@@ -68,6 +74,7 @@ class DeploymentProcessors(AbstractProcessorPackage):
         DeleteAutoScalingRuleAction, DeleteAutoScalingRuleActionResult
     ]
     create_access_token: ActionProcessor[CreateAccessTokenAction, CreateAccessTokenActionResult]
+    sync_replicas: ActionProcessor[SyncReplicaAction, SyncReplicaActionResult]
 
     def __init__(
         self, service: DeploymentServiceProtocol, action_monitors: list[ActionMonitor]
@@ -84,6 +91,7 @@ def __init__(
         self.create_deployment = ActionProcessor(service.create, action_monitors)
         self.destroy_deployment = ActionProcessor(service.destroy, action_monitors)
         self.create_access_token = ActionProcessor(service.create_access_token, action_monitors)
+        self.sync_replicas = ActionProcessor(service.sync_replicas, action_monitors)
 
     @override
     def supported_actions(self) -> list[ActionSpec]:
@@ -94,4 +102,5 @@ def supported_actions(self) -> list[ActionSpec]:
             UpdateAutoScalingRuleAction.spec(),
             DeleteAutoScalingRuleAction.spec(),
             CreateAccessTokenAction.spec(),
+            SyncReplicaAction.spec(),
         ]

src/ai/backend/manager/services/deployment/service.py

@@ -31,6 +31,10 @@
     DestroyDeploymentAction,
     DestroyDeploymentActionResult,
 )
+from ai.backend.manager.services.deployment.actions.sync_replicas import (
+    SyncReplicaAction,
+    SyncReplicaActionResult,
+)
 from ai.backend.manager.services.deployment.actions.update_auto_scaling_rule import (
     UpdateAutoScalingRuleAction,
     UpdateAutoScalingRuleActionResult,
@@ -136,3 +140,6 @@ async def create_access_token(
                 created_at=datetime.now(),
             )
         )
+
+    async def sync_replicas(self, action: SyncReplicaAction) -> SyncReplicaActionResult:
+        return SyncReplicaActionResult(success=True)