codex
33bf6e7f3d
These workplans exist only in the retired local hub. Their random pre-ADR-007 identifiers are refused by C-06 as stale references, so they cannot be registered. Deriving from the canonical record id takes no identity from anything: central does not hold them and the old ids die with the cache. Records central already holds were deliberately left untouched. Refs CUST-WP-0068-T06 Assistant: claude-code Assistant-Model: opus Assistant-Process: 2583210@bnt-lap001 Assistant-Session: f2bff2d5-e9b2-4338-92ca-10282a927006
4.4 KiB
| id | type | title | domain | repo | status | owner | topic_slug | created | updated | related | state_hub_workstream_id | ||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| RCLUSTER-WP-0014 | workplan | Publish cluster capacity, utilization, and allocation evidence | financials | railiance-cluster | finished | codex | railiance | 2026-08-11 | 2026-08-22 |
|
73ff6641-eaf6-5546-8f0e-b5aa1aa7f35b |
RCLUSTER-WP-0014 — cluster resource evidence
Goal
Publish reproducible non-secret evidence for
resource:railiance:reef-railiance:k3s: capacity, requested and observed
utilization, persistent-storage exposure, failure domains, and allocation
drivers. Origin: RESOURCE-WP-0003-T04 delegated evidence gap.
This is ADR-0006 reef evidence (capacity headroom, failure-domain
assumptions). It does not promote a binding to production-approved.
It does not publish a public metrics listener (ADR-0008). Collect
in-cluster or via the operator tunnel.
Current reality: one Ready node (railiance01), local storage, Traefik.
Do not invent a three-domain map or report Longhorn as live capacity.
RAIL-BS-WP-0007 is the HA target.
S2 publishes facts. reef-railiance cites them. resource-control /
fin-hub allocate. No booked EUR in this repo.
T01 — Produce periodic cluster capacity observations
id: RCLUSTER-WP-0014-T01
status: done
priority: high
state_hub_task_id: "abc4980e-4103-5d63-9e05-5a198dc7988b"
Collect node count, allocatable CPU/memory/storage/pods, workload requests and limits, observed usage when metrics are available, and PVC requested capacity. Aggregate by namespace/workload without exposing Secret data or pod environment. Preserve observation time, cluster identity, units, unavailable metrics, and source commands.
Done when repeated observations use a stable machine-readable interface and resource-control can distinguish headroom, reservations, and actual usage.
Outcome (2026-08-22): make cluster-observe REMOTE=railiance01 now emits
versioned, timestamped JSON with node allocatable/observed resources,
namespace/workload requests and limits, PVC/PV/storage-class state, stable
units, source commands, and explicit gaps without reading Secrets or pod
environments.
T02 — Publish the failure-domain and storage contract
id: RCLUSTER-WP-0014-T02
status: done
priority: high
state_hub_task_id: "9e0dc15e-a853-512e-be44-71525ce0bc47"
Expose the current node-to-host/provider failure-domain map and the actual
storage-class/PVC durability contract. Link the target ThreePhoenix work in
RAIL-BS-WP-0007 without reporting target nodes or distributed storage as
current capacity.
Done when consumers can calculate correlated-failure risk and recognize which
state remains local to railiance01.
Outcome (2026-08-22): the evidence contract reports one Ready node,
local-path storage, threephoenix_ready: false, and the missing independent
region/zone labels. Target HA and Longhorn capacity are not represented as
current facts.
T03 — Define technical allocation-driver evidence
id: RCLUSTER-WP-0014-T03
status: done
priority: medium
state_hub_task_id: "8a052726-e6f2-5852-aada-a14e67bca695"
Evaluate namespace/workload CPU-request hours, memory-request hours, PVC GB-month, and observed usage as allocation drivers. Publish numerator, denominator, residual/unallocated capacity, method version, and uncertainty; do not convert the result into booked spend.
Done when resource-control can create an explainable allocation and fin-hub can reconcile it to referenced financial facts.
Outcome (2026-08-22): cluster-raw-drivers-v1 publishes CPU requests,
memory requests, PVC bytes, observed usage, totals, residual capacity, and
uncertainty, while leaving formula selection and booked cost to
resource-control.
Acceptance
- Capacity evidence separates allocatable, requested, limited, and observed.
- Current single-host/local-storage risk is machine-readable.
- Allocation drivers reconcile to explicit totals and residuals.
Completion evidence
python3 -m unittest discover -s tests -p 'test_*.py' -v: 2 tests passed.- Live read-only capture:
docs/evidence/resource-reef-railiance-k3s/observations/20260822T100349Z.json. - Observed totals: 1 node, 53 workloads, 25 claims; 4,000m allocatable CPU, 3,830m requested CPU, 16,770,076,672 allocatable memory bytes, and 137,438,953,472 requested PVC bytes.