[Python] Worker was sent code 139! running Gunicorn + Chroma

I have a Flask app that uses Gunicorn/Nginx + ChromaDB v1.0.8. I've had no issues with its functionality till today when I restarted the service, to which I would receive

May 06 19:33:34 cluster gunicorn[6804]: 2025-05-06 19:33:34.413 | INFO | app:insert:79 - insert: querying chroma for nearest embeddings
May 06 19:33:34 cluster gunicorn[6802]: [2025-05-06 19:33:34 +0000] [6802] [ERROR] Worker (pid:6804) was sent code 139!
May 06 19:33:34 cluster gunicorn[6852]: [2025-05-06 19:33:34 +0000] [6852] [INFO] Booting worker with pid: 6852


This occurs when an endpoint is called to embed data into the vector database (Chroma), specifically when I go to query the collection for an embeddding

logger.info("insert: querying chroma for nearest embeddings")
query_result = collection.query(query_embeddings=[embedding], n_results=5)
distances = query_result["distances"][0]
metadatas_raw = query_result["metadatas"][0]
logger.info(f"insert: query returned {len(distances)} distances")


I've struggled to isolate the issue as ~1 out of 20 restarts will work, with the rest failing. I run Gunicorn with

ExecStart=/var/www/cluster/.venv/bin/gunicorn -k gthread --threads 4 --workers 1 --timeout 60 --bind unix:/var/www/cluster/cluster.sock -m 007 "app:app"


I can access the DB directly via the Python interpreter so the issue does not appear to be the database itself. Any ideas on what might be causing this?

Continue reading...