Recommendation for timeout setting

[farazhm]

Describe your question

When creating an eval I keep running into 504 Gateway Time-out. Is there any recommendation on what I should set as my timeout settings for my ELB

I've seen this https://langfuse.com/faq/all/self-hosting-502-504-network-errors and set my timeout to 2 minutes and it still fails, what is working for others?

Langfuse Cloud or Self-Hosted?

Self-Hosted

If Self-Hosted

3.148.0

If Langfuse Cloud

No response

SDK and integration versions

No response

Pre-Submission Checklist

• I have checked for existing issues/discussions and consulted Langfuse AI.

[inkeep-ai-assistant[bot]]

Based on the official documentation, the recommended approach for handling 502 and 504 network errors in self-hosted Langfuse is to configure your KEEP_ALIVE_TIMEOUT on the Langfuse Web container to be at least 5 seconds higher than your load balancer idle timeout.(1)

For example, if your ELB has an idle timeout of 120 seconds (2 minutes), you should set KEEP_ALIVE_TIMEOUT to at least 125 seconds on your Langfuse Web container.(1)

The reason for this configuration is that when a server closes a connection before the load balancer's idle timeout expires, the load balancer may attempt to reuse a connection it believes is still alive, which results in 502 errors.(1)

However, if you're experiencing 504 timeouts specifically during eval creation (rather than 502 errors), this suggests the request is taking longer to process than your load balancer timeout allows. In this case, you may need to:

1. Increase your ELB idle timeout beyond 2 minutes
2. Ensure KEEP_ALIVE_TIMEOUT remains at least 5 seconds higher than whatever ELB timeout you set(1)
3. Investigate backend performance issues that might be causing slow eval processing (check S3, Redis, ClickHouse, and worker container performance)(2)

The sources indicate that 504 timeouts during POST requests to the ingestion endpoint can be caused by backend bottlenecks involving S3 upload throttling, Redis queue delays, or worker/container resource exhaustion.(2)

📚 Sources:

1. Intermittent 502 and 504 network errors in self-hosted Langfuse
2. bug: Cant send traces, 504 Gateway Time-out but keep-alive is much higher than load balancer timeout

---

Have another question? Just tag @inkeep.

[farazhm]

figured out my issue, TLS wasn't enabled for redis, this post helped me out #11389