ansible-cluster

ansible-cluster is a collection of various Ansible roles pursuing the goal of an easy (HPC) cluster deployment. ansible-cluster does this by leveraging many open source technologies. A few are mentioned below:

• Ansible well, ansible-cluster is based on Ansible
• Apache as the webserver
• Cobbler as the install PXE server
• dnsmasq as the DHCP/DNS server for cluster nodes
• Slurm as the resource manager for the cluster

Disclaimer the docu is out of sync at the moment since the configuration has shifted towards default values and system variable loading

Idea of ansible-cluster

The idea behind ansible-cluster is that all relevant customizations and settings are done upfront and any command for handling a cluster operation happens via Ansible commands or applying some playbook, i.e. operating a (HPC) cluster is done in the Ansible context.

This means that the deployment of a (HPC) cluster should become as easy as

ansible-playbooks playbooks/clustersetup.yml

All an admin needs to customize or supply are the example and template files in the samples folder.

What does ansible-cluster do?

There are several roles implemented in this repository and some are external. A more detailed description is given in the roles section. ansible-cluster ansible-cluster is planned to support a RHEL 7.X based server as a frontend with maybe additional servers for storage. The compute and login nodes can be installed with Ubuntu (16.04), RHEL 7.X based distributions or Fedora (last known working Fedora was 24). Please have a look at the main config.

How to start off

To use ansible-cluster you need a RHEL/CentOS 7.X based system. Just grab any x86 64bit able system since ansible-cluster itself does not have heavy requirements. Slurm is a different matter though since a fast storage may be of interest.

Then grab first_start.sh from the ansible-cluster repository and change a few variables:

declare -A NEEDED_HOSTS=(["192.168.1.254"]="$(hostname) $(hostname -f) $(hostname -s)" ["192.168.1.15"]="needed_host_X")
REPO_HTTPS_LINK="https://github.com/AnKosteck/ansible-cluster.git"
#REPO_SSH_LINK="git@github.com:AnKosteck/ansible-cluster.git"
REPO_NAME="ansible-cluster"
HOSTS_FILE="cluster_hosts_example"
FORKS="5"

CLUSTER_ACCOUNTING="/root/ansible-cluster/samples/slurm_accounts_example.yml"
CLUSTER_CONFIG="/root/ansible-cluster/samples/config_template.yml"
CLUSTER_GROUPS="/root/ansible-cluster/samples/groups_example.yml"
CLUSTER_PARTITIONS="/root/ansible-cluster/samples/partition_config_example.yml"
CLUSTER_SSH_LIMITS="/root/ansible-cluster/samples/ssh_access_limits_example.yml"
CLUSTER_USERS="/root/ansible-cluster/samples/users_example.yml"

supply your own configurations (please look at the samples) and execute the script via ./first_start.sh This script sets up Ansible and Git, then clones the ansible-cluster Repository and finishes off by calling

ansible-playbook playbooks/first_start/first_start.yml

From then on, the cluster can be managed with Ansible and ansible-cluster. first_start.yml sets up a second interface for internal services (PXE, DHCP, ...) and disables SELinux. After the frontend hast been restarted, ansible-cluster should be usable.

Templates and examples

In the samples folder examples and templates are given for you to start with.