Added prod start script, tweaked swarm values for inference containers

compose/dev/docker-compose.yml

@@ -4,12 +4,11 @@ services:
     build:
       context: ../../
       dockerfile: compose/dev/django/Dockerfile
-    env_file:
-      - ../../.env
+    env_file: ../../.env
     volumes:
       - ../../:/app
     ports:
-      - '0.0.0.0:8000:8000'
+      - "8000:8000"
     depends_on:
       fyp-postgres-dev:
         condition: service_healthy
@@ -23,27 +22,26 @@ services:
       dockerfile: compose/dev/node/Dockerfile
     environment:
       NODE_ENV: development
-      CHOKIDAR_USEPOLLING: 'true'
+      CHOKIDAR_USEPOLLING: "true"
     stdin_open: true
     volumes:
       - ../../site:/app:delegated
       - /app/node_modules
     ports:
-      - '0.0.0.0:5173:5173'
+      - "5173:5173"
 
   fyp-postgres-dev:
     container_name: fyp-postgres-dev
     image: pgvector/pgvector:pg15
-    env_file:
-      - ../../.env
+    env_file: ../../.env
     environment:
       POSTGRES_HOST_AUTH_METHOD: trust
     volumes:
       - fyp_postgres_data:/var/lib/postgresql/data
     ports:
-      - '0.0.0.0:5432:5432'
+      - "5432:5432"
     healthcheck:
-      test: ['CMD-SHELL', 'pg_isready -h 127.0.0.1 -p 5432 -U ${POSTGRES_USER} -d ${POSTGRES_DB}']
+      test: ["CMD-SHELL", "pg_isready -h 127.0.0.1 -p 5432 -U $${POSTGRES_USER} -d $${POSTGRES_DB}"]
       interval: 5s
       timeout: 3s
       retries: 5
@@ -52,7 +50,7 @@ services:
     container_name: fyp-redis-dev
     image: redis:7-alpine
     ports:
-      - '0.0.0.0:6379:6379'
+      - "6379:6379"
     volumes:
       - fyp_redis_data:/data
     healthcheck:
@@ -66,8 +64,7 @@ services:
     build:
       context: ../../
       dockerfile: compose/dev/celery/Dockerfile
-    env_file:
-      - ../../.env
+    env_file: ../../.env
     volumes:
       - ../../:/app
     depends_on:
@@ -76,21 +73,17 @@ services:
       fyp-postgres-dev:
         condition: service_healthy
 
-
   fyp-inference-dev:
     container_name: fyp-inference-dev
     build:
       context: ../../
       dockerfile: compose/dev/inference/Dockerfile
-    env_file:
-      - ../../.env
+    env_file: ../../.env
     volumes:
       - ../../:/app
       - ../../models:/app/models
       - hf_cache:/root/.cache/huggingface
     deploy:
-      mode: replicated
-      replicas: 1
       resources:
         reservations:
           devices:
@@ -104,14 +97,13 @@ services:
       - HF_HOME=/root/.cache/huggingface
       - HF_HUB_OFFLINE=1
     ports:
-      - "0.0.0.0:8001:8001"
+      - "8001:8001"
     depends_on:
       fyp-redis-dev:
         condition: service_healthy
       fyp-postgres-dev:
         condition: service_healthy
 
-
 volumes:
   fyp_postgres_data:
   fyp_redis_data:

compose/prod/django/start

@@ -0,0 +1,27 @@
+#!/bin/bash
+
+set -o errexit
+set -o pipefail
+set -o nounset
+
+DB_HOST="${POSTGRES_HOST}"
+DB_PORT="${POSTGRES_PORT}"
+
+echo "Waiting for database at ${DB_HOST}:${DB_PORT}..."
+wait-for-it ${DB_HOST}:${DB_PORT} --timeout=30 --strict || {
+	echo "Timed out waiting for database" >&2
+	exit 1
+}
+
+echo "Database is available, continuing startup..."
+
+python manage.py makemigrations
+python manage.py migrate --noinput
+
+for fixture in /app/data/*.json; do
+    echo "Loading fixture: $fixture"
+    python manage.py loaddata "$fixture"
+done
+
+python manage.py collectstatic --noinput
+exec daphne -b 0.0.0.0 -p 8000 config.asgi:application

compose/prod/docker-compose.inference.yml

@@ -6,8 +6,6 @@ services:
       dockerfile: compose/dev/inference/Dockerfile
     restart: unless-stopped
     deploy:
-      mode: replicated
-      replicas: 1
       resources:
         reservations:
           devices:
@@ -20,6 +18,10 @@ services:
       - INFERENCE_HTTP_HOST=0.0.0.0
       - INFERENCE_HTTP_PORT=8001
       - NVIDIA_VISIBLE_DEVICES=all
+      - WATCHFILES_FORCE_POLLING=true
+      - PYTHONPATH=/app
+      - HF_HOME=/root/.cache/huggingface
+      - HF_HUB_OFFLINE=1
     ports:
       - '0.0.0.0:58001:8001'
     volumes: