frame_id: /systems/cloud-sim-infrastructure
Six manual steps into one deployment
Turned a slow, error-prone manual Isaac Sim provisioning process on cloud GPUs into a single reproducible Docker Compose deployment.
Problem
Every new cloud GPU instance for Isaac Sim development required manually pulling repositories, running setup steps, installing dependencies, and wiring paths inside the container by hand — slow enough that developers avoided using cloud simulation as their default workflow.
Investigation
Built a setup script that provisions the YAML + Dockerfile + startup script chain automatically, clones and updates the required simulation repositories, runs the simulation repo's setup script, and mounts a shared volume for embedded `.so` files reused by the Isaac Sim loader and simulation tooling. Along the way, diagnosed and resolved two separate blocking production issues: GPU instances stuck cycling between building/running/unhealthy states requiring manual intervention, and the Isaac Sim streaming client GUI silently failing to load because a service was bound to the wrong port — traced through errors that only surfaced on `docker compose down`.
Outcome
A fresh cloud GPU instance now goes from zero to a working Isaac Sim development environment through one reproducible deployment, with the repo-clone, dependency, and streaming pipeline all verified working end to end.