Connection is dropped early with Content-Type "application/stream+json"

[sylvain-rouquette]

Describe the bug
Connection is dropped early with Content-Type "application/stream+json", could be ambassador or AWS ELB.

To Reproduce
I have a K8S cluster deployed in AWS with kops, and ambassador is in front of my java backend. I also have an auth service in Go connected to ambassador.
Everything works fine with application/json requests.
I implemented an endpoint with Flux to support application/stream+json and text/event-stream. The endpoint works fine locally (without ambassador), but when I deploy it in my cluster and try a curl, I get this error.
"curl: (18) transfer closed with outstanding read data remaining"
I receive some data (2400 json objects over 10000)
When I check the logs, I notice that ambassador logs twice:

• ambassador service receives the request
• my java service receive the request
• another ambassador service receives the request

The connection is closed when the second ambassador service receives the request.
Why does it receive that second request, could it be related?

Here is the response headers:

< HTTP/1.1 200 OK
< content-encoding: gzip
< expires: 0
< cache-control: no-cache, no-store, max-age=0, must-revalidate
< x-xss-protection: 1; mode=block
< pragma: no-cache
< x-frame-options: DENY
< date: Thu, 07 Feb 2019 15:38:31 GMT
< x-content-type-options: nosniff
< content-type: application/stream+json
< x-envoy-upstream-service-time: 11
* Server envoy is not blacklisted
< server: envoy
< transfer-encoding: chunked

Expected behavior
The connection shouldn't be closed before I receive all the data.

Versions (please complete the following information):

• Ambassador: 0.50.0, but didn't work on 0.40.2 as well.
• Kubernetes environment AWS
• Version 1.10.11

Additional context
I need help, but it seems I can't but the tag on this issue.

[sylvain-rouquette]

I was able to fix my problem by setting a timeout in my mapping. The hint was in ambassador's log, 'x-envoy-expected-rq-timeout-ms', '3000'.

Is this the expected behavior to have the request closed by ambassador, while there's still some data being downloaded?

My request doesn't have a Content-Length since it's a stream, it could be related.

Feel free to close my bug if you consider it isn't one.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.