Avoid the load balancing Serverless endpoints (not ready for Beta)

Details:
  • There's an undocumented 2MB payload limit
  • Request randomly takes 1 seconds to 120 seconds (cut off), despite being under 2MB.
  • It randomly get's either a Cloudflare 504 (the request times out), or Cloudflare 502 Bad gateway
  • Even if you have set workers, it frequently returns "error":"no workers available"
  • It randomly cuts connections and drops payloads
  • Some servers randomly get offline for a period of time (still printing 200 in the log), but no external connection
  • I've tried a production setting and simple test cases, nothing works reliably
  • It's not a case of one faulty instance, the entire load balancing set-up is unreliable
  • This is all with more than one active worker
This is what logs look when using <0.5MB image input data:
Screenshot_2025-10-31_at_16.50.51.png
Was this page helpful?