← ClaudeAtlas

sales-demos-probe-envlisted

Measure what a cluster actually has — allocatable, what is already requested, what is genuinely free — and emit a recommended available_memory_gb instead of trusting a hardcoded one. Strictly read-only, so it is safe mid-demo. Runs playbooks/probe_env.yml. TRIGGER when: the user asks how much room a cluster has, whether an add-on or another VM will fit, why Terraform refuses a tier, or has moved to a new or resized RHDP environment and the memory budget may be stale. SKIP: if the user is asking whether a VM build would be fast — that is ocpvirt-new-env — or if OpenShift Virtualization is not installed at all, which is ocpvirt-setup.
ericcames/sales.demos · ★ 1 · AI & Automation · score 67
Install: claude install-skill ericcames/sales.demos
# sales-demos-probe-env ## There is an AAP path now too (#330) `Cluster Day 0 - Probe Capacity` runs this same playbook from AAP. It carries a `read-only` label — the one thing the other label axes could not express — so it is filterable as safe to run during a live demo, which is exactly when someone needs to ask whether another VM will fit. Answers one question: **what does this cluster actually have, and what will fit on it?** ## Why this exists `terraform/ocpvirt/variables.tf` carried `available_memory_gb = 14` as a hardcoded guess. It was measured once, on a smaller cluster, and then outlived it — sandbox now has roughly **five times** that free. Nothing reported the drift, and nothing could have. The budget guard in `locals.tf` fails *closed*: an under-provisioned figure does not error, it quietly refuses tiers the cluster could run easily. A demo gets smaller and nobody learns why. The next RHDP environment will differ again. A hardcoded number is exactly what produced the stale one, so this is re-runnable rather than a paragraph in an issue (#100). ## It is strictly read-only Every task is `k8s_info`. It creates nothing, deletes nothing, and a run reports `changed=0`. **Safe to run in the middle of a live demo** when someone asks whether the cluster can take another VM. That is the whole reason it is a second playbook rather than a flag on `prepare_env.yml`, which builds and destroys a real VM to do its job. ## Requests, not usage — the distinction that mat