I downloaded the Enterprise Nucleus Server - Compose Stack from the NVIDIA NGC Catalog, version nucleus-stack-2023.2.8+tag-2023.2.8.gitlab.24971896.fea9b67c, and completed the installation by following the documentation here: Installing an Enterprise Nucleus Server.
In the same network, there are two computers, A and B. On computer A, I installed the Nucleus Compose Stack. When connecting from computer B to the Nucleus Server on A, everything works fine. However, when trying to connect from computer A to its own Nucleus Server, the connection fails.
Specifically, from computer B, I can connect to the Nucleus Server via applications like Isaac Sim or USD Composer, and I can also access the Nucleus Navigator using A’s IP:8080. But on computer A itself, neither of these methods works.
By default, Docker containers may not expose their services to “localhost” on the host when using bridge networking.
From another computer (B), accessing via A’s external IP (e.g., A_IP:8080) works, but localhost:8080 or 127.0.0.1:8080 on A may not route traffic correctly to the container.
Solution: In your docker-compose.yml, set network_mode: host for the Nucleus service. This change ensures the server inside the container binds directly to all interfaces—including localhost—on the host.
For both computers A and B, you need to access the Nucleus server though the hostname IP you’ve assigned within the base_stack.env
When you say you cannot connect to from computer A to its own Nucleus Server, what exactly do you mean? You are using the full IP or base stack defined hostname, correct?
@Richard3D is correct that using 127.0.0.1 or localhost will not work in this case.
It would be good to know what errors you see on computer A, especially if trying to connect through Navigator.