Skip to content

Latest commit

 

History

History
115 lines (81 loc) · 3.18 KB

OPENSTACK.md

File metadata and controls

115 lines (81 loc) · 3.18 KB
Raw

ansible-role-slurm

Build Status

Creates a slurm cluster

Tested with slurm versions:

  • 14.11.0
  • 14.11.3
  • 15.08.x

Tested with these linux distributions:

  • CentOS 6
  • Only 14.11.x
  • CentOS 7
  • Only 15.08.x

How-To

Initial playbook configuration

Playbook variables:

All variables should be defined in defaults/main.yml

All nodes run munge. Nodes which are part of the slurm_compute host group will additionally run slurmd. Nodes which are part of the slurm_service host group will additionally runs slurmctld and slurmdbd.

You also need to add a mysql_slurm_password: "PASSWORD" string somewhere. This will be used to set a password for the slurm mysql user.

Below instructions are for running slurm and perhaps multiple slurm instances in openstack

These instructions no longer work, files have been removed. A playbook that uses this role: https://github.com/fgci-org/fgci-ansible

Initial configuration of the instances and your workstation:

  • First yum -y install nc on the bastion host.
  • Then setup SSH config so you don't have to have a public IP on each instance. Change the Hostname in "Host bastion" to the service node.

Put this in ~/.ssh/config :

# http://edgeofsanity.net/article/2012/10/15/ssh-leap-frog.html
# This applies to all hosts in your ssh config.
ControlMaster auto
ControlPath ~/.ssh/ssh_control_%h_%p_%r

# Always ssh as cloud-user and don't save hostkeys
Host bastion
  User cloud-user
  StrictHostKeyChecking no
  UserKnownHostsFile /dev/null
  Hostname 86.50.168.39
     
Host slurm* 
   User cloud-user
   StrictHostKeyChecking no
   UserKnownHostsFile /dev/null
   ForwardAgent yes
   ProxyCommand ssh bastion nc %h %p
  • Second update the /etc/hosts on the service node (playbook should update the others):
# For second cluster
192.168.36.126 slurm2-compute1
192.168.36.125 slurm2-service
192.168.36.124 slurm2-login

# For first cluster
192.168.36.129 slurm-compute1
192.168.36.128 slurm-service
192.168.36.127 slurm-login

Then we can finally run the slurm configuration playbooks:

Description of the playbooks:

  • site.yml - calls the slurm*.yml playbooks
  • slurm_*.yml # The playbooks that configure the servers
  • set slurm_version - this is used to determine which version to download from schedmd.com

Run them in this order:

  • Update stage to have the right IP addresses and hostnames.

  • configuring 1st slurm: ansible-playbook site.yml

  • Update stage to have the right IP addresses and hostnames.

  • configuring 2st slurm: ansible-playbook site.yml

Add cloud-user to slurm

sacctmgr create account name=csc
sacctmgr create user name=cloud-user account=csc

Make changes to slurm.conf and distribute it to nodes and restart/reconfigure:

  • Would be nice with a role /tag where one could just run ansible-playbook site.yml --tag new-slur-config and it pushes new config and restarts/reconfigs as necessary.

Authors / Contributors: