Merge pull request #5 from RC-CHN/bay-dashboard

Bay dashboard

Author: RC-CHN

.gitignore

@@ -185,3 +185,4 @@ logs/
 
 pkgs/bay/ship_data/
 pkgs/bay/scripts/
+pkgs/bay/tests/k8s/k8s-deploy-local.yaml

pkgs/bay/Dockerfile

@@ -1,27 +1,84 @@
+# ============================================
+# Stage 1: Build frontend (Vue.js Dashboard)
+# ============================================
+FROM node:22-alpine AS frontend-builder
+
+WORKDIR /app/dashboard
+
+# Copy package files first for better caching
+COPY dashboard/package.json dashboard/package-lock.json ./
+
+# Install dependencies
+RUN npm ci
+
+# Copy source files
+COPY dashboard/ ./
+
+# Build for production
+RUN npm run build
+
+# ============================================
+# Stage 2: Build Python dependencies
+# ============================================
+FROM python:3.11-slim AS python-builder
+
+WORKDIR /app
+
+# Install build dependencies
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    gcc \
+    libc6-dev \
+    && rm -rf /var/lib/apt/lists/*
+
+# Copy requirements and install dependencies
+COPY requirements.txt ./
+RUN pip wheel --no-cache-dir --wheel-dir /app/wheels -r requirements.txt
+
+# ============================================
+# Stage 3: Final image with Nginx + Python
+# ============================================
 FROM python:3.11-slim
 
-# Install system dependencies
-# RUN apt-get update && apt-get install -y \
-#     gcc \
-#     libc6-dev \
-#     libffi-dev \
-#     bash \
-#     && rm -rf /var/lib/apt/lists/*
+# Install Nginx and curl (for health checks)
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    nginx \
+    curl \
+    && rm -rf /var/lib/apt/lists/*
 
-# Set working directory
 WORKDIR /app
 
-# Copy all project files
-COPY pyproject.toml requirements.txt alembic.ini run.py ./
+# Copy Python wheels and install
+COPY --from=python-builder /app/wheels /app/wheels
+COPY requirements.txt ./
+RUN pip install --no-cache-dir --no-index --find-links=/app/wheels -r requirements.txt \
+    && rm -rf /app/wheels
+
+# Copy Python application files
+COPY pyproject.toml alembic.ini run.py ./
 COPY app/ ./app/
 COPY alembic/ ./alembic/
 
-# Install dependencies and create data directory in one layer
-RUN pip install -r requirements.txt --no-cache-dir && \
-    mkdir -p /app/data
+# Create data directory
+RUN mkdir -p /app/data
+
+# Copy built frontend to Nginx html directory
+COPY --from=frontend-builder /app/dashboard/dist /usr/share/nginx/html
+
+# Copy Nginx configuration
+COPY nginx.conf /etc/nginx/nginx.conf
+
+# Copy and prepare entrypoint script
+COPY docker-entrypoint.sh /docker-entrypoint.sh
+RUN chmod +x /docker-entrypoint.sh
+
+# Expose ports:
+# - 8156: API (Python backend, can be exposed publicly)
+# - 8157: Dashboard (Nginx, can be hidden behind NAT)
+EXPOSE 8156 8157
 
-# Expose port
-EXPOSE 8156
+# Health check
+HEALTHCHECK --interval=30s --timeout=10s --start-period=10s --retries=3 \
+    CMD curl -f http://localhost:8157/nginx-health && curl -f http://localhost:8156/health || exit 1
 
-# Start the application
-CMD ["python", "run.py"]
+# Start both services
+ENTRYPOINT ["/docker-entrypoint.sh"]

pkgs/bay/app/database.py

@@ -3,7 +3,7 @@
 from sqlalchemy.pool import StaticPool
 from typing import Optional, List
 from app.config import settings
-from app.models import Ship, SessionShip
+from app.models import Ship, SessionShip, ShipStatus
 from datetime import datetime, timezone
 
 
@@ -63,10 +63,12 @@ async def update_ship(self, ship: Ship) -> Ship:
         ship.updated_at = datetime.now(timezone.utc)
         session = self.get_session()
         try:
-            session.add(ship)
+            # Use merge() instead of add() to handle detached objects
+            # merge() copies the state of the given instance into a persistent instance
+            merged_ship = await session.merge(ship)
             await session.commit()
-            await session.refresh(ship)
-            return ship
+            await session.refresh(merged_ship)
+            return merged_ship
         finally:
             await session.close()
 
@@ -87,10 +89,23 @@ async def delete_ship(self, ship_id: str) -> bool:
             await session.close()
 
     async def list_active_ships(self) -> List[Ship]:
-        """List all active ships"""
+        """List all active ships (running and creating)"""
+        session = self.get_session()
+        try:
+            # Include both RUNNING and CREATING status ships
+            statement = select(Ship).where(
+                (Ship.status == ShipStatus.RUNNING) | (Ship.status == ShipStatus.CREATING)
+            )
+            result = await session.execute(statement)
+            return list(result.scalars().all())
+        finally:
+            await session.close()
+
+    async def list_all_ships(self) -> List[Ship]:
+        """List all ships (including stopped)"""
         session = self.get_session()
         try:
-            statement = select(Ship).where(Ship.status == 1)
+            statement = select(Ship).order_by(Ship.created_at.desc())
             result = await session.execute(statement)
             return list(result.scalars().all())
         finally:
@@ -163,20 +178,21 @@ async def update_session_ship(self, session_ship: SessionShip) -> SessionShip:
         """Update session-ship relationship"""
         session = self.get_session()
         try:
-            session.add(session_ship)
+            # Use merge() instead of add() to handle detached objects
+            merged_session_ship = await session.merge(session_ship)
             await session.commit()
-            await session.refresh(session_ship)
-            return session_ship
+            await session.refresh(merged_session_ship)
+            return merged_session_ship
         finally:
             await session.close()
 
     async def find_available_ship(self, session_id: str) -> Optional[Ship]:
         """Find an available ship that can accept a new session"""
         session = self.get_session()
         try:
-            # Find ships that have available session slots
+            # Find ships that have available session slots (only RUNNING ships)
             statement = select(Ship).where(
-                Ship.status == 1, Ship.current_session_num < Ship.max_session_num
+                Ship.status == ShipStatus.RUNNING, Ship.current_session_num < Ship.max_session_num
             )
             result = await session.execute(statement)
             ships = list(result.scalars().all())
@@ -193,38 +209,50 @@ async def find_available_ship(self, session_id: str) -> Optional[Ship]:
             await session.close()
 
     async def find_active_ship_for_session(self, session_id: str) -> Optional[Ship]:
-        """Find an active running ship that this session has access to"""
+        """Find an active running ship that this session has access to.
+        
+        If the session has access to multiple running ships, returns the most recently updated one.
+        """
         session = self.get_session()
         try:
-            # Find active ships that this session has access to
+            # Find RUNNING ships that this session has access to
+            # Order by updated_at desc to get the most recently used one
             statement = (
                 select(Ship)
                 .join(SessionShip, Ship.id == SessionShip.ship_id)
                 .where(
                     SessionShip.session_id == session_id,
-                    Ship.status == 1,
+                    Ship.status == ShipStatus.RUNNING,
                 )
+                .order_by(Ship.updated_at.desc())
             )
             result = await session.execute(statement)
-            return result.scalar_one_or_none()
+            # Use scalars().first() instead of scalar_one_or_none() to handle multiple results
+            return result.scalars().first()
         finally:
             await session.close()
 
     async def find_stopped_ship_for_session(self, session_id: str) -> Optional[Ship]:
-        """Find a stopped ship that belongs to this session"""
+        """Find a stopped ship that belongs to this session.
+        
+        If the session has access to multiple stopped ships, returns the most recently updated one.
+        """
         session = self.get_session()
         try:
-            # Find stopped ships that this session has access to
+            # Find STOPPED ships that this session has access to
+            # Order by updated_at desc to get the most recently stopped one
             statement = (
                 select(Ship)
                 .join(SessionShip, Ship.id == SessionShip.ship_id)
                 .where(
                     SessionShip.session_id == session_id,
-                    Ship.status == 0,
+                    Ship.status == ShipStatus.STOPPED,
                 )
+                .order_by(Ship.updated_at.desc())
             )
             result = await session.execute(statement)
-            return result.scalar_one_or_none()
+            # Use scalars().first() instead of scalar_one_or_none() to handle multiple results
+            return result.scalars().first()
         finally:
             await session.close()
 
@@ -266,5 +294,88 @@ async def decrement_ship_session_count(self, ship_id: str) -> Optional[Ship]:
         finally:
             await session.close()
 
+    async def delete_sessions_for_ship(self, ship_id: str) -> List[str]:
+        """Delete all session-ship relationships for a ship and return deleted session IDs"""
+        session = self.get_session()
+        try:
+            # First, get all session IDs for this ship
+            statement = select(SessionShip).where(SessionShip.ship_id == ship_id)
+            result = await session.execute(statement)
+            session_ships = list(result.scalars().all())
+            
+            deleted_session_ids = [ss.session_id for ss in session_ships]
+            
+            # Delete all session-ship relationships
+            for ss in session_ships:
+                await session.delete(ss)
+            
+            await session.commit()
+            return deleted_session_ids
+        finally:
+            await session.close()
+
+    async def extend_session_ttl(
+        self, session_id: str, ttl: int
+    ) -> Optional[SessionShip]:
+        """Extend the TTL for a session by updating expires_at"""
+        from datetime import timedelta
+        
+        session = self.get_session()
+        try:
+            statement = select(SessionShip).where(SessionShip.session_id == session_id)
+            result = await session.execute(statement)
+            session_ship = result.scalar_one_or_none()
+
+            if session_ship:
+                now = datetime.now(timezone.utc)
+                session_ship.expires_at = now + timedelta(seconds=ttl)
+                session_ship.last_activity = now
+                session.add(session_ship)
+                await session.commit()
+                await session.refresh(session_ship)
+
+            return session_ship
+        finally:
+            await session.close()
+
+    async def expire_sessions_for_ship(self, ship_id: str) -> int:
+        """Mark all sessions for a ship as expired by setting expires_at to current time.
+        
+        This is called when a ship is stopped to ensure session status
+        reflects the actual container state.
+        
+        Args:
+            ship_id: The ship ID
+            
+        Returns:
+            Number of sessions updated
+        """
+        session = self.get_session()
+        try:
+            statement = select(SessionShip).where(SessionShip.ship_id == ship_id)
+            result = await session.execute(statement)
+            session_ships = list(result.scalars().all())
+            
+            now = datetime.now(timezone.utc)
+            updated_count = 0
+            
+            for ss in session_ships:
+                # Only update if session is still active (expires_at > now)
+                expires_at = ss.expires_at
+                if expires_at is not None:
+                    if expires_at.tzinfo is None:
+                        expires_at = expires_at.replace(tzinfo=timezone.utc)
+                    if expires_at > now:
+                        ss.expires_at = now
+                        session.add(ss)
+                        updated_count += 1
+            
+            if updated_count > 0:
+                await session.commit()
+            
+            return updated_count
+        finally:
+            await session.close()
+
 
 db_service = DatabaseService()

pkgs/bay/app/main.py

@@ -6,7 +6,7 @@
 from app.database import db_service
 from app.drivers import initialize_driver, close_driver
 from app.services.status import status_checker
-from app.routes import health, ships, stat
+from app.routes import health, ships, stat, sessions
 
 # Configure logging
 logging.basicConfig(
@@ -85,6 +85,7 @@ def create_app() -> FastAPI:
     app.include_router(health.router, tags=["health"])
     app.include_router(ships.router, tags=["ships"])
     app.include_router(stat.router, tags=["stat"])
+    app.include_router(sessions.router, tags=["sessions"])
 
     return app
 

pkgs/bay/app/models.py

@@ -5,10 +5,18 @@
 import uuid
 
 
+# Ship status constants
+class ShipStatus:
+    """Ship status constants"""
+    STOPPED = 0   # Ship is stopped, container not running
+    RUNNING = 1   # Ship is running, container active
+    CREATING = 2  # Ship is being created, container not yet ready
+
+
 # Database Models
 class ShipBase(SQLModel):
     id: str = Field(default_factory=lambda: str(uuid.uuid4()), primary_key=True)
-    status: int = Field(default=1, description="1: running, 0: stopped")
+    status: int = Field(default=ShipStatus.CREATING, description="0: stopped, 1: running, 2: creating")
     created_at: datetime = Field(
         default_factory=lambda: datetime.now(timezone.utc),
         sa_column=Column(DateTime(timezone=True)),
@@ -87,6 +95,10 @@ class CreateShipRequest(BaseModel):
     max_session_num: int = Field(
         default=1, gt=0, description="Maximum number of sessions that can use this ship"
     )
+    force_create: bool = Field(
+        default=False,
+        description="If True, skip all reuse logic and always create a new container"
+    )
 
 
 class ShipResponse(BaseModel):
@@ -127,6 +139,12 @@ class ExtendTTLRequest(BaseModel):
     ttl: int = Field(..., gt=0, description="New TTL in seconds")
 
 
+class StartShipRequest(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    ttl: int = Field(default=3600, gt=0, description="TTL in seconds for the started ship")
+
+
 class ErrorResponse(BaseModel):
     detail: str