[Error]Failed to create a new stream

[brano.pejar]

Hi!
I successfully deployed the Omniverse Kit App Streaming stack on AWS following the official NVIDIA deployment guide.
Using the Omniverse Web Viewer sample, I attempted to initialize a new streaming session.
However, the process fails during the stream loading phase.
Upon investigation, I identified the following error in the streaming service pod logs:

{“Timestamp”: 1761760928616654523, “SeverityNumber”: 17, “SeverityText”: “ERROR”, “InstrumentationScope”: “”, “Body”: "Failed to start a new stream: <class 'KeyError'>, 'mediaPort'", “Attributes”: {“app.clock”: 6452.414436102, “timestamp”: “2025-10-29T18:02:08.616654523Z”, “correlation_id”: “8b67222d82d04d4e9be9fd14577e325c”}, “Resource”: {“app.namespace”: “nv.svc”, “app.name”: “nv.svc.streaming”, “app.instance.id”: “90156564-c6b6-4229-9113-a8bea0eb7601”, “app.version”: “1.11.0”}}
{“Timestamp”: 1761760956985732732, “SeverityNumber”: 9, “SeverityText”: “INFO”, “InstrumentationScope”: “”, “Body”: "Failed to connect to ``omni-streaming-nlb-3a92083aabbd5ac6.elb.us-east-1.amazonaws.com:447``. Connection not ready.", “Attributes”: {“app.clock”: 6480.783531666, “timestamp”: “2025-10-29T18:02:36.985732732Z”, “correlation_id”: “6da20f255d764e7ca4154c7aaea3c8c4”}, “Resource”: {“app.namespace”: “nv.svc”, “app.name”: “nv.svc.streaming”, “app.instance.id”: “90156564-c6b6-4229-9113-a8bea0eb7601”, “app.version”: “1.11.0”}}

How can I resolve this issue?

[Richard3D]

You cannot directly view an Omniverse kit stream from an AWS server, over the public internet, back to your local machine, if that is what you are trying to. At least, is not set up that way, by default. You would have to engineer the stack to be publicly available over WLAN and the internet. You would have to edit the ports and the access to make it work.

The first thing to check is, can you access the stream, on the same remote AWS machine as the stack, using a normal web browser? Then can you access the stream on another AWS machine, configured to read across the internet AWS stack. Then finally, you would have to config it to be “open access”, which can be risky.