My autoscaled instance fails when running "git clone" on a private repo.

I do have the SSH key placed at /root/.ssh/id_rsa on the machine, and when I SSH into the machine and run sudo su; git clone <the repo> it succeeds.

Also, in the extra_vm_bash_script field: I added a whoami command which prints root , so it seems like the user being used to run the git clone during task execution is in fact root .

For context, here's the startup command that the autoscaler runs:

python -m clearml_agent --config-file /root/clearml.conf daemon --queue aws_4gpu_machines --docker python:3.9

Full log included...