Unexpected 500 direct_response when Service has zero Endpoints (should it be 503?)

[yuntanghsu]

Description:

What issue is being seen? Describe what should be happening instead of
the bug, for example: The expected value isn't returned, etc.

When the only Pod behind a Service goes down and the Service ends up with zero Endpoints, Envoy Gateway returns:

500 Internal Server Error
response_code_details: "direct_response"
upstream_cluster: null

I expected Envoy Gateway to return 503 Service Unavailable (no healthy upstream)

I would like to confirm whether returning 500 direct_response in this scenario is the intended behavior.

Repro steps:

Include sample requests, environment, etc. All data and inputs
required to reproduce the bug.

1. Deploy a Service with a single Pod.
2. Create an HTTPRoute pointing to that Service.
   The route is valid (Accepted=True, ResolvedRefs=True).
3. Send a request → routing works normally.
4. Delete or crash the only Pod.
5. Service now has 0 Endpoints.
6. Send the same request again.
7. Envoy Gateway returns 500 direct_response, and Envoy does not attempt any upstream connection (upstream_cluster=null).

Note: If there are privacy concerns, sanitize the data prior to
sharing.

Environment:

Include the environment like gateway version, envoy version and so on.

envoyproxy-gateway:v1.5.0
envoy:v1.34.1

Logs:

Include the access logs and the Envoy logs.

Envoy access log
{
":authority":"aaa",
"method":"POST",
"protocol":"HTTP/2",
"response_code":500,
"response_code_details":"direct_response",
"upstream_cluster":null,
"upstream_host":null,
"x-envoy-origin-path":"ddd"
}

HttpRoute status:
status:
parents:

• conditions:
  • type: Accepted
    status: "True"
    reason: Accepted
  • type: ResolvedRefs
    status: "True"
    reason: ResolvedRefs

Is this behavior expected?
Should Envoy Gateway return 503 instead of generating a 500 direct_response when a backend Service has zero Endpoints?

[zirain]

@zhaohuabing I recalled there's a fix recently?