Isaac-Sim+OIGE fails with container automation

I was trying to deploy on AWS using terraform and in the terraform resource “aws_instance” user_data = file("") cannot properly run the script “”. The weird behaviour is that sometimes it runs, sometimes it runs partially and sometimes dont run at all.
The behaviour might be because of execution time, so I was thinking to implement Ansible. I would like to know what other alternatives you reccomend or if Nvidia has Ansible playbooks to setup its images, would be useful to come with a robust strategy to deploy OIGE at scale.

resource "aws_instance" "isaac_sim_oige" {
  # ami             = local.nvidia_ami
  ami             = data.aws_ami_ids.nvidia_omniverse_ami.ids[0]
  instance_type   = local.instance_type
  key_name        = "isaac-sim-oige-key"
  user_data	      = file("")
  # user_data	      = file("")
  security_groups = [ ]

  # We Gonna pick the first availability zone that has the Instance Type we want
  availability_zone = keys({ for az, details in data.aws_ec2_instance_type_offerings.my_ins_type :
  az => details.instance_types if length(details.instance_types) != 0 })[0]

  subnet_id       =

  # Env = "isaac-sim"
  ebs_block_device {
    device_name = "/dev/sda1"
    volume_size = 200

  tags = {
    Name = local.instance_name
    Env = var.env

  depends_on = [

The entire code can be found here:

Hi. We currently do not have a Terraform or Ansible playbook to deploy Isaac Sim.
To reduce execution time, you could upload the shader cache folder to the AMI.

You could try running a Cloudformation template for the Linux VDI. See Deploy an Omniverse Virtual Workstation on AWS (Ubuntu Linux 20.04) — Omniverse Microservices documentation


No worries, I made a solution.
I would love if we could clean it up and post it on Nvidia Blog.

Maybe, make it part of the documentation.

I also hit lots of issues with OV farm, thats the next step. Isaac Sim on K8s

1 Like

Sure. Sounds good. We can help with that.


@Pappachuck_renan Thank you for your offer to add this to our documentation, that’s very nice of you.
We appreciate your contribution.