fix(ops): rebalance container memory limits to prevent celery OOM kills

Celery worker was OOM-killed (41 restarts) at 512MB with 4 concurrent
workers. Reduce concurrency to 2, increase worker limit to 768MB, and
reclaim memory from over-provisioned services (db 512→256, beat 256→128,
flower 256→192). Total allocation stays within 4GB server budget.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

app/core/celery_config.py

@@ -91,7 +91,7 @@ celery_app.conf.update(
     task_soft_time_limit=25 * 60,  # 25 minutes soft limit
     # Worker settings
     worker_prefetch_multiplier=1,  # Disable prefetching for long tasks
-    worker_concurrency=4,  # Number of concurrent workers
+    worker_concurrency=2,  # Keep low on 4GB servers to avoid OOM
     # Result backend
     result_expires=86400,  # Results expire after 24 hours
     # Retry policy

docker-compose.yml

@@ -11,7 +11,7 @@ services:
       - postgres_data:/var/lib/postgresql/data
     ports:
       - "5432:5432"
-    mem_limit: 512m
+    mem_limit: 256m
     healthcheck:
       test: ["CMD-SHELL", "pg_isready -U orion_user -d orion_db"]
       interval: 30s
@@ -86,7 +86,7 @@ services:
     volumes:
       - ./logs:/app/logs
       - ./exports:/app/exports
-    mem_limit: 512m
+    mem_limit: 768m
     healthcheck:
       test: ["CMD-SHELL", "celery -A app.core.celery_config inspect ping --timeout 10 || exit 1"]
       interval: 30s
@@ -107,7 +107,7 @@ services:
     depends_on:
       redis:
         condition: service_healthy
-    mem_limit: 256m
+    mem_limit: 128m
     healthcheck:
       disable: true
     networks:
@@ -128,7 +128,7 @@ services:
     depends_on:
       redis:
         condition: service_healthy
-    mem_limit: 256m
+    mem_limit: 192m
     healthcheck:
       test: ["CMD-SHELL", "curl -f http://localhost:5555/ || exit 1"]
       interval: 30s