Omnia collections

Ansible Collections Structure

Ansible collections are organized by the following naming convention:

namespace.collection.[role|module]

For Omnia's purposes, the namespace will be omnia.

Omnia collections

The omnia namespace should contain collections which are grouped by function:

• cluster - a collection of playbooks for executing the creation, destruction, and resizing of clusters
• appliance - a collection of playbooks for bootstrapping an Omnia provisioning/deployment appliance
• provision - a collection of roles for deploying various provisioning tools
• accelerator - a collection of roles for detecting and deploying various hardware accelerator drivers and software stacks
• wlm - a collection of roles for deploying various workload managers
• storage - a collection of roles for deploying storage appliances and solutions
• network - a collection of roles for deploying networking infrastructure and software-defined networks
• monitoring - a collection of roles for deploying various monitoring tools
• logging - a collection of roles for deploying various logging tools
• platform - a collection of roles for deploying various user-facing platforms
• management - a collection of roles for deploying management tools and dashboards

omnia.cluster

Playbooks

• create
• destroy
• grow
• shrink

omnia.appliance

Playbooks

• create

omnia.provision

Roles

• cobbler (default)
• warewulf (not implemented)
• maas (not implemented)
• openstack (not implemented)
• idrac

omnia.accelerator

Playbooks

• deploy - deploy supported accelerators

Roles

• detect - a role for installing custom fact gathering script and refreshing gather_fact
• nvidia - a role for installing Nvidia GPU drivers and deploying software/services
• amd - a role for installing AMD GPU drivers and deploying software/services
• xilinx - a role for installing Xilinx FPGA drivers and deploying software/services
• alterra - a role for installing Intel FPGA drivers and deploying software/services
• intel - a role for installing Intel Xe GPU drivers and deploying software/services

omnia.wlm

Playbooks

• slurm - deploy Slurm clusters by leveraging slurm_manager and slurm_worker roles
• kubernetes - deploy Kubernetes clusters by leveraging k8s_manager and k8s_worker roles

Roles

• slurm_manager
• slurm_worker
• k8s_manager
• k8s_worker

omnia.storage

Playbooks

• beegfs - deploy a BeeGFS filesystem leveraging beegfs_mds and beegfs_ost roles
• lustre - deploy a Lustre filesystem leveraging lustre_mds and lustre_ost roles

Roles

• me4 - configuring Dell EMC PowerVault ME4 arrays
• me5 - configuring Dell EMC PowerVault ME5 arrays
• onefs - configuring Dell EMC Isilon/PowerScale appliances and deploying CSI operators
• beegfs_mds
• beegfs_ost
• lustre_mds
• lustre_ost
• nfs - deploy a NFS filesystem

omnia.network

Roles

• os10 - configure Dell EMC PowerSwitch Ethernet switch running OS10
• sonic - configure ethernet switch running SoNIC
• infiniband - deploy an Infiniband fabric
• roce - deploy a RoCE fabric
• calico - deploy calico SDN
• flannel - deploy flannel SDN

omnia.monitoring

Roles

• prometheus
• ganglia

omnia.logging

Roles

• fluentd
• elk

omnia.platform

Roles

• kubeflow
• jupyterhub
• open_ondemand

omnia.management

Roles

• lens
• k8s_dashboard
• awx