ToolCall request returns 504 Gateway Timeout via Nginx Ingress for Long-Running Requests

[CeerDecy]

Problem Description

We observe intermittent 504 errors when handling long-running requests through Nginx Ingress, despite the SSE backend functioning properly. Below are the key details:

Environment

1. Backend Service:

• mcp-server deployed in Kubernetes
• Used server.WithKeepAliveInterval(10 * time.Second) for SSE connections
• Processing logic requires 2+ minutes to complete

2. Gateway Layer:

• Nginx Ingress Controller default timeout settings (60s)

Observed Behavior

client by python-sdk:

• Successful results via /sse endpoint.
• the ToolCall request receive 504 Gateway Timeout after 60s, though backend logs confirm continued processing.

client by mcp-go:

• timeout, no result return

[Hwting]

Problem Description

We observe intermittent 504 errors when handling long-running requests through Nginx Ingress, despite the SSE backend functioning properly. Below are the key details:

Environment

1. Backend Service:

• mcp-server deployed in Kubernetes
• Used server.WithKeepAliveInterval(10 * time.Second) for SSE connections
• Processing logic requires 2+ minutes to complete

2. Gateway Layer:

• Nginx Ingress Controller default timeout settings (60s)

Observed Behavior

client by python-sdk:

• Successful results via /sse endpoint.
• the ToolCall request receive 504 Gateway Timeout after 60s, though backend logs confirm continued processing.

**client by `mcp-go`:**

• timeout, no result return

@CeerDecy How is SSE URL customization implemented? Why can't I modify it and only write the IP address? How to deploy a remote MCP server and access it through FQDNS?

[CeerDecy]

solved