fix OOM kills by reducing worker counts, cuts memory utilization (#406)

Author: irvingpop

Dockerfile

@@ -61,7 +61,6 @@ RUN --mount=type=cache,target=/var/cache/apt,sharing=locked \
     apt-get update && apt-get install -y --no-install-recommends \
     libpq5 \
     curl \
-    wget \
     && apt-get upgrade -y
 
 # Create non-root user for security
@@ -110,7 +109,6 @@ RUN --mount=type=cache,target=/var/cache/apt,sharing=locked \
     apt-get update && apt-get install -y --no-install-recommends \
     libpq5 \
     curl \
-    wget \
     && apt-get upgrade -y
 
 # Create non-root user for security
@@ -121,7 +119,8 @@ RUN groupadd -r appuser && \
 ENV PYTHONUNBUFFERED=1 \
     PYTHONPATH=/app/src \
     PATH="/app/.venv/bin:$PATH" \
-    VIRTUAL_ENV=/app/.venv
+    VIRTUAL_ENV=/app/.venv \
+    DJANGO_Q_WORKERS=1
 
 WORKDIR /app
 

src/gunicorn_config.py

@@ -64,13 +64,18 @@
 #       A positive integer. Generally set in the 1-5 seconds range.
 #
 
-workers = 3
+workers = 1
+threads = 2
 worker_class = "gthread"
 worker_connections = 1000
 timeout = 30
 keepalive = 2
 worker_tmp_dir = "/dev/shm"
 
+# Memory leak mitigation - restart workers after handling N requests
+max_requests = 1000
+max_requests_jitter = 50
+
 #   preload_app - Load application code before forking worker processes.
 #       This conserves memory and speeds up server boot times by loading
 #       the Django application once in the main process, then forking

src/settings/components/base.py

@@ -104,7 +104,7 @@
 # Django-Q2 configuration (sync mode for testing, use redis/orm in production)
 Q_CLUSTER = {
     "name": "operationcode",
-    "workers": 2,
+    "workers": config("DJANGO_Q_WORKERS", default=1, cast=int),
     "timeout": 60,
     "retry": 120,
     "queue_limit": 50,