Skip to content

Instantly share code, notes, and snippets.

@sunhwan
Last active April 11, 2023 19:48
Show Gist options
  • Save sunhwan/9faafdc37cd29c6b8d16f8689e618cc0 to your computer and use it in GitHub Desktop.
Save sunhwan/9faafdc37cd29c6b8d16f8689e618cc0 to your computer and use it in GitHub Desktop.
AWS parallel-cluster config file
[aws]
aws_region_name = us-east-1
[aliases]
ssh = ssh {CFN_USER}@{MASTER_IP} {ARGS}
[global]
cluster_template = default
update_check = true
sanity_check = true
[cluster default]
key_name = sunhwan-2021
base_os = alinux2
scheduler = slurm
ebs_settings = default
master_instance_type = t2.medium
vpc_settings = default
queue_settings = gpu, cpu
[vpc default]
vpc_id = vpc-0ce86ec9e4e89bdda
master_subnet_id = subnet-02d741c1794baddb8
compute_subnet_id = subnet-099c460314207b832
use_public_ips = true
[ebs default]
shared_dir = /shared
volume_type = sc1
volume_size = 500
[queue gpu]
enable_efa = false
compute_resource_settings = gpu
compute_type = spot
[queue cpu]
enable_efa = false
compute_resource_settings = cpu
compute_type = spot
[compute_resource gpu]
instance_type = p3.2xlarge
min_count = 0
initial_count = 0
max_count = 10
spot_price = 1.5
[compute_resource cpu]
instance_type = c5.2xlarge
min_count = 0
initial_count = 0
max_count = 10
spot_price = 0.5
@jubilantsundar
Copy link

Very useful post. When I used this config file, the /shared folder only has 20GB instead of 500GB.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment