No-Tears-Cluster CLI command

[sean-smith]

The value in no-tears-cluster comes from the ability to make the pcluster config file generic, i.e. you can share a pcluster config file by posting it on S3 and anyone can use it just by setting ConfigS3Uri to point to that file.

This feature request proposes decoupling that from the stack creation via a CLI command.

$ no-tears-cluster configure cluster_template.ini -o cluster.ini
$ pcluster create cluster -c cluster.ini

Where cluster.ini could look like:

[global]
cluster_template = hpc
update_check = true
sanity_check = true

[aws]
aws_region_name = ${AWS_DEFAULT_REGION}

[aliases]
ssh = ssh {CFN_USER}@{MASTER_IP} {ARGS}

[cluster hpc]
key_name = ${ssh_key_id}
base_os = ubuntu1804
scheduler = slurm
master_instance_type = c5.2xlarge
compute_instance_type = c5.18xlarge
vpc_settings = public-private
fsx_settings = fsx-scratch2
disable_hyperthreading = true
dcv_settings = dcv
post_install = ${post_install_script_url}
post_install_args = "/shared/spack-0.15 releases/v0.15 /opt/slurm/log sacct.log"
s3_read_resource = arn:aws:s3:::*
s3_read_write_resource = ${s3_read_write_resource}/*
initial_queue_size = 0
max_queue_size = 10
placement_group = DYNAMIC
master_root_volume_size = 50
compute_root_volume_size = 30
ebs_settings = myebs
cw_log_settings = cw-logs

[ebs myebs]
volume_size = 200
shared_dir = /shared

[fsx fsx-scratch2]
shared_dir = /scratch
storage_capacity = 1200
deployment_type = SCRATCH_2
import_path=${s3_read_write_url}

[dcv dcv]
enable = master
port = 8443
access_from = 0.0.0.0/0

[cw_log cw-logs]
enable = false

[vpc public-private]
vpc_id = ${vpc_id}
master_subnet_id = ${master_subnet_id}
compute_subnet_id = ${compute_subnet_id}
use_public_ips = false

[rsignell-usgs]

@sean-smith, I'm getting back into no-tears-cluster and this sounds awesome.