Cluster

All Ansible source code, playbooks, roles, and testing are centralized on <strong>stargate.research.bitsmasher.net</strong>:

• <strong>Primary workspace</strong>: \texttt{~/workspace/lab-franklin/ansible/} on stargate
• <strong>Secondary (chonk)</strong>: \texttt{/mnt/backup1/workspace/lab-franklin/ansible/} (mirror for local builds)
• <strong>Interpreter</strong>: \texttt{/home/franklin/.local/bin/ansible} on stargate
• <strong>Testing standard</strong>: Native \texttt{ansible-test} (sanity + integration) -- Molecule deprecated as of August 2026

The lab uses cluster-wide automation via ansible-playbook targeting inventory groups defined in ansible/hosts}. Cluster management is handled through role-based provisioning rather than a dedicated orchestration layer.

k3s Clustershell (Historical)

Previous cluster shell groupings remain documented for reference:

all     -- all managed nodes
head    -- control plane nodes
compute -- worker nodes
gpu     -- GPU-enabled Jetson nodes

These groupings are now managed through Ansible inventory patterns rather than clustershell.

Centralized Ansible Workflow

1. Role changes committed to stargate workspace
2. Native ansible-test runs on stargate (zero external cost)
3. Playbook execution targets remote hosts via SSH batch probes
4. Changes verified via dynamic reachability: \texttt{ssh -o BatchMode=yes -o ConnectTimeout=3 franklin@\textless host\textgreater}