This repo contains ansible playbooks extracted from elasticluster's playbooks for installing a slurm based cluster on CLIMB, with few modification.
As CLIMB is built on top of OpenStack, ideally, we can use elasticluster to install slurm conveniently. However, the OpenStack API on CLIMB is likely not available, I eventually ended up with this workaround.
- Install ansible:
pip install ansible - Login to Horizon [url not shown here] and create a your own private key pair. Download the key to you local computer
- Initiate manually instances using Launch a custom server from https://bryn.climb.ac.uk/. The cluster should have 1 head and at least 1 node.
- Clone this repo:
git clone https://github.com/thanhleviet/ansible-slurm-climb - Copy the key pair into the folder
ansible-slurm-climb - Create an inventory file, define server name and ip for the head and the nodes. See the
hosts.cardifffor example. - Replace location to the key pair and inventory in
ansible.cfg - To install the cluster, run this command from a terminal:
ansible-playbook climb.yml
The playbook climb.yml will install slurm for a e cluster of 1 head and a numbers of nodes defined in hosts.cardiff . As I used the cluster temporarily for analyses, I removed some setup which may take the installation longer (i.e. nis) and destroy all the instances after I finished. This playbook also install conda, nextflow, singularity, docker for running my own bioinformatics pipelines.
Tested with Ubuntu 16.04 distro but the playbook should work with Debian/Centos/REHL and even on physical computers
Thanks MRC for this valuable computing resource.