2026-05-09 — v6 Fleet Install Day

The single biggest operational day on the v6 fleet — hub-dc-v6 and spoke-dc-v6 both installed, ODF data plane brought up on the spoke, and the management pull-model baselined. The implicit day-wrap that sits between the pre-v6 reset and the day-2 wave.

2026-05-09 is the single most consequential operational day in the v6 fleet’s lifetime. Both clusters — hub-dc-v6 and spoke-dc-v6 — went from agent ISOs on ocp-bootstrap to oc get nodes green inside ~16 hours. The day did not get an explicit “day wrap” GitHub issue the way 2026-05-10 did under #159; the implicit day-wrap is the mid-day session-closeout report at opp-full-plat/reports/sessions/20260509-112323-session-closeout.md, plus the chain of ~30 per-session reports the day produced.

This page collects what happened that day, what shape the fleet took, and where to pick up the thread.

Fleet shape that emerged

Reading the diagram:

• Hub (left) is the compact 3-AIO management cluster — three nodes, each carrying control-plane,master,worker roles. No physical workers, no ODF, deliberately storage-light per ADR 0018.
• Spoke (right) is the workload cluster — three VM masters plus three physical worker boxes (gold-1, gold-2, gpu-01), with the physical workers running on the 30.30.0.0/16 NIC ports after the first MAC selection picked the wrong path.
• Green dashed line is the ACM pull model: the spoke registers to the hub, the hub manages the spoke through ACM + per-cluster Argo, and platform-gitops is the source of truth on both sides.

The hub install completed at 08:00 UTC; the spoke base install completed at 15:20 UTC. Everything else on this day was the work needed to get those two milestones across the line, plus the immediate day-2 baseline that turned the installed clusters into a fleet.

The implicit day-wrap

reports/sessions/20260509-112323-session-closeout.md (closed at 11:23 UTC, mid-day) is the day’s only explicit closeout artifact. It scoped the morning:

The user did approve, mid-afternoon, and the spoke went up. The afternoon’s work isn’t covered by that closeout — it’s covered by the per-session report chain.

Per-session report chain — what happened, in order

Times are UTC. Slugs map to files under opp-full-plat/reports/sessions/20260509-<HHMMSS>-<slug>.md.

Morning — hub install + image supply

The hub install itself was the easy part — the real load-bearing piece of work was the catalog baseline at 08:43, which is what made the hub usable for operator installs. That manifest set is exactly what REPRO-HUB-1 later adopted into platform-gitops (see the dedicated REPRO-HUB-1 page).

Mid-morning — developer readiness + federated GitOps ADRs

The morning split into two threads. While the hub was rolling its catalog baseline, a second thread set up the developer-readiness and federated-GitOps contracts:

By 10:08 the hub was Argo-bootstrapped against platform-gitops and the federated-GitOps contract was written. From this point, every subsequent operator install on either cluster ships through a GitLab MR.

Mid-day — spoke install package + workspace governance

Afternoon — spoke base install + ODF disk reset

The spoke base install itself was clean. ODF was the painful part — the disks weren’t safe to wipe without a review pass, the first attempt left residue, and the cleanup ran across 16:16-16:27 before the cluster was ready for a GitOps-driven ODF install.

Evening — ACM registration + Nexus image-supply baseline

By 23:27 the fleet had the foundational guardrails written down: ADR 0014 (developer-readiness), ADR 0015 (federated GitOps), ADR 0017 (parallel-agent worktree isolation), ADR 0018 (pull-model baseline), ADR 0019 (Nexus-only image supply). Every operator install that ships after this day is bound by them.

What this day produced

How this day relates to the rest of the history

This day sits between two other entries in the history-and-replay section:

• Upstream: pre-v6 → v6 transition. The pre-v6 fleet (hub-dc, spoke-dc, hub-dr, spoke-dr under ADR 0001) was retired; ADR 0022 supersedes ADR 0001 and reserves the v6 names. 2026-05-09 is the day the active half of that supersession became physical clusters.
• Downstream: the day-2 wave. Everything after 2026-05-09 — operator installs, PCI-DSS chain, ODF storage-consumer cleanup, ACM gitops-addon Routes-CRD incident, REPRO-HUB-1 adoption — runs against the clusters this day produced. The Day-Wrap Archive carries the per-day record from 2026-05-10 onward.

Why the implicit day-wrap was enough

A formal day-wrap GitHub issue (the way 2026-05-10 got #159) wasn’t needed for this day. The two install milestones each have their own session-closeout records (hub-dc-v6-boot-install-complete, spoke-dc-v6-install-complete) with the actual Install complete lines, the oc get nodes output, the residual risks, and the next-step gates. The mid-day session-closeout covers the morning’s ADR + milestone work. The afternoon and evening run was per-session reports plus SESSION_LOG.md. Together they produce the same audit trail as a day-wrap roll-up, with less duplication.

The convention going forward (set on 2026-05-05 18:39 UTC under session-routines.md): per-session reports are mandatory for any non-trivial session; day-wrap issues are reserved for days that produce enough roll-up-able churn that the per-session reports alone leave gaps. 2026-05-09 didn’t have those gaps — it had two crisp install completes and the contract ADRs to bind them.

References

• Session reports (UTC times): all under opp-full-plat/reports/sessions/20260509-<HHMMSS>-<slug>.md
• Day’s implicit closeout: opp-full-plat/reports/sessions/20260509-112323-session-closeout.md
• Hub install completion: opp-full-plat/reports/sessions/20260509-080024-hub-dc-v6-boot-install-complete.md
• Spoke install completion: opp-full-plat/reports/sessions/20260509-152004-spoke-dc-v6-install-complete.md
• Hub catalog baseline (becomes REPRO-HUB-1 input): opp-full-plat/reports/sessions/20260509-084300-hub-dc-v6-disconnected-catalog-baseline.md
• Pre-v6 to v6 transition — what came before
• REPRO-HUB-1 hub adoption — how the hub day-1 baseline became GitOps after the fact
• Day-Wrap Archive — the per-day record from 2026-05-10 onward
• ADRs accepted this day: 0014 (developer-readiness), 0015 (federated GitOps), 0017 (parallel-agent worktree isolation), 0018 (pull-model baseline), 0019 (Nexus-only image supply)