All about StormShift Automation

Ansible based, of course! (and a litle bit of gitops)

Cluster as a Service via Ansible Automation Platform

We want to provide

• Standalone OpenShift Cluster on OpenShift Virtualization Control plan as VM ) Issue: create automation for ocp cluster mno #76
  • From Single-Node over Compact to full-blown OpenShift Clusters.
• Hosted Control Plane (HCP) Cluster with KubeVirt/OpenShift-Virt provider
• RHEL for Edge / MicroShift deployments

Standalone / HCP Cluster - Design

• Playbooks should provide a ready to use cluster.
  • No addtional ACM/GitOps integration is needed, but optional possible
• Core playbooks/roles provide a cluster and the cluster can enrich with feature roles that are executed during installation. Example for features:
  • kubevirt-csi driver configuration and deployment roles/feature-kubevirt-csi
  • Adding cluster to central ACM instance roles/feature-manage-with-acm
  • ...

Infrastructure to host the clusters

• Bare-Metal OpenShift Virtualization Cluster - called managment (mgmt.) cluster
• Mgmt. Cluster provides various storage backends
  • Red Hat OpenShift Data Foundation
  • NetApp Trident
  • LVMS (For HCP etcd only)
  • ...
• Every cluster running at the mgmt. cluster get his own namespace/project like that: stormshift-<cluster-name>-infra for example stormshift-ocp11-infra
• At the mgmt. Cluster there will be group stormshift-admins this group gets admin privileges on all stormshift related projects.
• The last person who deployed a cluster gets cluster admin on the cluster-specific namespace as well.
• All clusters & nodes are predefined in a ansible inventory: inventory/hosts.yml - more details look into Ansible inventory description

Ansible inventory description

The inventory is managed in the inventory/ folder.

• Every cluster is a ansible group cluster_<cluster-name> for example and as subgroups:

  • cluster_<cluster-name>_cp for control plane
  • cluster_<cluster-name>_workers for worker nodes

• Every node in a cluster is part of the group.

• Every node start with the <cluster-name>-control-1 for example:

  • ocp11-control-1
  • ocp11-control-2
  • ocp11-control-3
  • ocp11-worker-1
  • ...

    inventory
    ├── group_vars
    │   ├── all.yaml
    │   ├── cluster_ocp11.yaml
    ...
    ├── host_vars
    │   ├── ocp11-control-1.yaml
    └── hosts.yml

Ansible Automation Platform Configuration

• Create stormshift organisation
• Create stormshift-admin team, make team to stormshift org. admin/full-access
• Add execution environment to stormshift organisation
• Add project stormshift-automation with this repo to the stormshift organisation as well.
• Add stormshift-inventory inventory
  • Add source automation-repo based on Source from project:
    • Important - Inventory file: inventory/hosts.yml
• Deploy stormshift

Ansible playbook and role structure/idea

• One playbook to rule them all:
  • deploy-cluster.yaml
  • destroy-cluster.yaml
• Playbook is splitted into 3 steps aka plays:
  • pre-deploy - runs against hosts[0] - only the first one
  • deploy instances/vm's - runs against all hosts to create and start all VM's
  • post-deploy - runs against hosts[0] - only the first one
• Main role called cluster
  • Main tasks files:
    • pre-deploy-cluster-{{ cluster_type (sno|classic) }}.yaml
    • deploy-cluster-{{ cluster_type (sno|classic) }}.yaml
    • post-deploy-cluster-{{ cluster_type (sno|classic) }}.yaml

Run playbooks local

# add all informations
cp -v development-example.vars development-example.vars-private
vim development-example.vars-private
cp -v development-example.env development-example.env-private
vim development-example.env-private

# Load env variables
source development-example.env-private

# Run it
ansible-navigator run ./deploy-cluster.yaml -e @development-example.vars-private -v

Ansible inventory structure (folder: inventory/)

Documentation:

• https://docs.ansible.com/ansible/latest/inventory_guide/intro_inventory.html#organizing-host-and-group-variables

How do dump the entire inventory locally:

ansible-navigator run dump-inventory.yaml --list-hosts

How to dump vars of a host locally:

ansible-navigator run dump-inventory.yaml --limit common-pattern*

• common-pattern: https://docs.ansible.com/ansible/latest/inventory_guide/intro_patterns.html#common-patterns

How to build execution env.

podman login registry.redhat.io
export VERSION=$(date +%Y%m%d%H%M)

ansible-builder build \
    --verbosity 3 \
    --container-runtime podman \
    --tag quay.coe.muc.redhat.com/stormshift/automation-execution-environment:$VERSION

podman login quay.coe.muc.redhat.com
podman push quay.coe.muc.redhat.com/stormshift/automation-execution-environment:$VERSION