How to add tags/label for billing purposes and segregate users access with HPC Slurm?

[edl4006]

I'm looking to test and use HPC Slurm through the cluster-toolkit but have some confusion with this particular things, hoping these are easy questions to be answered.

1. When I spin up more instances from my Slurm , how can I add a unique tag so I can segregate cost and billing associated to the nodes that was just spun up in the GCP Billing console through Slurm?
2. When I spin up instances via Slurm I want users to be able to access their own pool of nodes, is there a way to set up a specific service accounts access when I provision via Slurm or ensure that I can segregate users access somehow based off of slurm provisioning?

Thank you.

[tpdownes]

Hi @edl4006

For billing purposes we advise adding labels to your Slurm nodeset (the schedmd-slurm-gcp-v6-nodeset module) with the following input variable:

https://github.com/GoogleCloudPlatform/cluster-toolkit/blob/main/community/modules/compute/schedmd-slurm-gcp-v6-nodeset/README.md#input_labels

Each nodeset is automatically labeled with ghpc_deployment: deployment_name, where the deployment_name matches the value in your blueprint.

If I understand your 2nd question, you can assign custom service accounts to each nodeset following our "enterprise" example":

cluster-toolkit/examples/hpc-enterprise-slurm.yaml

Lines 74 to 111 in f9f9256

	- id: compute_sa
	source: community/modules/project/service-account
	settings:
	name: compute
	project_roles:
	- logging.logWriter
	- monitoring.metricWriter
	- storage.objectCreator
	- id: homefs
	source: modules/file-system/filestore
	use: [network]
	settings:
	local_mount: /home
	- id: projectsfs
	source: modules/file-system/filestore
	use: [network]
	settings:
	local_mount: /projects
	# This file system has an associated license cost.
	# https://console.developers.google.com/marketplace/product/ddnstorage/exascaler-cloud
	- id: scratchfs
	source: community/modules/file-system/DDN-EXAScaler
	use: [network]
	settings:
	local_mount: /scratch
	- id: n2_nodeset
	source: community/modules/compute/schedmd-slurm-gcp-v6-nodeset
	use: [network, compute_sa]
	settings:
	node_count_dynamic_max: 4
	machine_type: n2-standard-2
	instance_image: $(vars.slurm_image)
	enable_placement: false # the default is: true
	allow_automatic_updates: false

We do not have any examples of access control on Slurm partitions. Is your goal to create groups of POSIX users who can access Partition A with Service Account A and Partition B with Service Account B?