Discussion on: How to automate creating high end virtual machines on AWS for data science projects

View post

Hi! Thank for your post. You may try Sparrowform as provision tool, this might some reduce Bash complexity and let you run provision independently on Terraform:

So instead of provisioner "remote-exec" ... inside your terrafrom scenario just ( scenario for CentOS7 ):

$ cat aws_instance.Node.sparrowfile

for (
    'git', 'vim-enhanced.x86_64',
    'python36u', 'python36u-pip', 'python36u-devel'
  ) -> $p {
  package-install $p
}

bash "pip3.6 install jupyter";

my $user = input_params('SshUser');

directory "/home/$user/.jupyter", %( owner => $user );


file "/home/$user/.jupyter/jupyter_notebook_config.py", %(
  owner    => $user,
  content  => "c.NotebookApp.allow_origin = '*'
c.NotebookApp.ip = '0.0.0.0'"
)

And then:

$ terraform apply
$ sparrowform --ssh_user=centos --ssh_private_key=~/Downloads/alexey-test.pem