Repository navigation
epic(cluster): hardware inventory & first-node selection #1080
Description
Hardware inventory & first-node selection
Purpose
Know exactly what hardware exists, in what condition, so the first 3-5 cluster nodes are picked on evidence instead of memory.
Scope
Gather full hardware stats for all ~12 Threadripper machines (CPU, board, RAM, every NVMe/HDD with model + SMART, NICs, GPUs). Note the 8 HDDs still in the box (no baseline data exists yet). Rank machines by identical-ness and drive quality and name the first 3-5 cluster nodes and their order.
Execution mode
Hybrid: the inventory design, templates, and analysis are online; gathering stats needs machines powered and reachable (physical), or BMC/KVM access once cabled.
Dependencies
Feeds every other epic. Node selection gates epics 5 and 6.
Acceptance
A committed inventory (one record per machine) covering every machine and drive; a named, ordered first-node list with the reasoning recorded; open questions (exact NIC spec, switch gear) either answered or filed as task issues.
Structure
Part of the basement cluster program. Project home: planning/roadmaps/basement-cluster.md in this repo. Full program register: goal workspace files/idea-register-2026-10-20.md (Muse goal goal_ba17a5777c21). Detailed task issues are filed as sub-issues of this epic; dependencies use blocks / blocked-by links.
Hard rules for the whole program: no HashiCorp anything; everything bootstrappable from config, no hand-built state.
Activity
developing-today commentedon Oct 27, 2026
CODE-222
dezren39 commentedon Oct 27, 2026
Epic map — basement cluster program
This epic: #1080 — Hardware inventory & first-node selection
Dependencies: Feeds all others; gates #1085, #1086
Filed in the 2026-10-26 cluster seed. Project home:
planning/roadmaps/basement-cluster.md.docs(basement-cluster): link filed epic issues in roadmap
1 remaining item
dezren39 commentedon Nov 3, 2026
agent-note:v1
state: triaged (coverage pass)
Coverage pass 2026-11-03: reviewed in the fleet-wide per-issue pass (drift / overlap / already-handled / stale / blockers / ordering). Verdict: still live; queued in repo-health-code-pilot as triage at priority 30 (priority/none). If this is already handled at current main, the triage closes it with evidence; if it needs splitting, the triage files the split.
Found via: coverage pass 2026-11-03
dezren39 commentedon Nov 3, 2026
agent-note:v1
state: picked-up → blocked (escalated to Drew — physical access required)
queue: repo-health-code-pilot
task: 7624 (blocked, reason class
escalated-to-user)by: openchamber-scheduled-lane
at: 2026-11-03T14:00Z
Verdict: still live, and not workable from this repo — the blocker is hardware time
Revalidated at current
main(4c75cc808):planning/roadmaps/basement-cluster.md:13still carries this epic as a planned row — "Gather hardware stats for all ~12 machines and pick the first 3–5 cluster nodes" — with no status change since the roadmap was written.planning/roadmaps/holds only the roadmap, the electrician brief and the repo roadmaps; aThreadrippersearch across tracked Markdown returns the roadmap and the electrician brief, no per-machine records. So the epic's acceptance (one committed record per machine and drive, a named ordered first-node list) is not quietly done.Why this is Drew's call, not a worker's
The epic's own execution mode is
hybrid, and the missing half is physical: "gathering stats needs machines powered and reachable (physical), or BMC/KVM access once cabled." A code-queue worker cannot power twelve machines, cable BMC, or read SMART data off drives that are still in the box (the 8 HDDs the epic explicitly says have no baseline). Everything a worker could contribute — the record schema, the analysis templates — is already described as online, and making those up front without a single real record would be busywork against a format nobody has validated.So the exact decision required:
Handling
escalated-to-user), issue labelleduser/escalatedin place ofstatus/needs-triage;queue/blocked+status/blockedset,work/assignedkept while it waits.planning/roadmaps/(or wherever epic(cluster): hardware inventory & first-node selection #1080 names), then produce the ranked first-node list.