Slurm Workload Manager#
Simple Linux Utility for Resource Management (Slurm) is a workload resource manager that is used to schedule job items or tasks on a cluster of systems. Slurm is comprised of three components a Controller, Compute Nodes and Clients. These guides cover setting up a virtual machine to host the Controller and other guides for deploying the required components on the User/Compute nodes where clients will submit and execute their jobs.
Operating Systems#
These guides are written for Red Hat Enterprise Linux 8 based operating systems and are compatible with AlmaLinux 8 and Rocky Linux 8.
Access Controls#
Standard cluster user accounts must be able to SSH into the Slurm Controller node in order for Slurm to operate correctly.
Hostnames#
These guides use the following example Hostnames and FQDN for Slurm Controller and Slurm Compute and Client nodes. FQDNs can be provided by utilizing a DNS server or installing a hosts file on the system itself.
Note
An example /etc/hosts file has been provided: hosts
Controller#
The Slurm Controller is deployed on the following node:
Hostname |
FQDN |
Function |
Type |
IPv4 Address |
|---|---|---|---|---|
slurm |
slurm.engwsc.example.com |
Slurm Controller |
VM Guest |
192.168.1.82 |
Compute/Client#
The Slurm Compute and Client agents are deployed on the following nodes:
Hostname |
FQDN |
Function |
Type |
IPv4 Address |
|---|---|---|---|---|
user01 |
user01.engwsc.example.com |
User Node 1 |
Bare-Metal |
192.168.1.60 |
user02 |
user02.engwsc.example.com |
User Node 2 |
Bare-Metal |
192.168.1.61 |
user03 |
user03.engwsc.example.com |
User Node 3 |
Bare-Metal |
192.168.1.62 |
user04 |
user04.engwsc.example.com |
User Node 4 |
Bare-Metal |
192.168.1.63 |
comp01 |
comp01.engwsc.example.com |
Compute Node 1 |
Bare-Metal |
192.168.1.64 |
comp02 |
comp02.engwsc.example.com |
Compute Node 2 |
Bare-Metal |
192.168.1.65 |
comp03 |
comp03.engwsc.example.com |
Compute Node 3 |
Bare-Metal |
192.168.1.66 |
comp04 |
comp04.engwsc.example.com |
Compute Node 4 |
Bare-Metal |
192.168.1.67 |
Prerequisites#
In order for Slurm to operate correctly, IdM and NFS must already be in place to handle user authentication and NFS based home directories. Password-less SSH’ing is also a requirement for Slurm but is configured after the User/Compute nodes are deployed.
Guides#
The following guides are required to deploy the Slurm Controller:
The following guides are required for deploying the Slurm Compute and Clients:
The following are Slurm usage guides: