core-hub/workplans/CORE-WP-0007-haskell-retirement.md

195 lines
7.3 KiB
Markdown
Raw Normal View History

---
id: CORE-WP-0007
type: workplan
title: "Retire Haskell/IHP infrastructure after Core Hub cutover"
domain: infotech
repo: core-hub
status: active
owner: codex
topic_slug: custodian
created: "2026-06-27"
updated: "2026-07-08"
state_hub_workstream_id: "a8d66822-e435-4b1e-ad81-37a298d1795e"
---
# Retire Haskell/IHP infrastructure after Core Hub cutover
Remove the Haskell dependency from the production interaction framework after the replacement is proven.
`CORE-WP-0005` finished 2026-07-03: `hub.coulomb.social` ingress now serves Core
Hub; Inter-Hub compatibility, staging import, dual-run smokes, and production
cutover gates are closed. Haskell retirement no longer waits on migration work —
it waits on a short post-cutover stabilization window, operator approval to
retire the Inter-Hub rollback deployment, and the four tasks below.
## Retire Probe Repo
```task
id: CORE-WP-0007-T01
status: done
priority: medium
state_hub_task_id: "d3a494a3-dc23-491f-af32-1d19120b3aae"
```
Rename or archive `ihp-railiance-probe` (`coulomb/ihp-railiance-probe`),
preserving `INTENT.md` and `DeploymentBlueprint.md` notes about the failed
IHP/GHC/Nix build path. `IRP-WP-0001` is already `done`; this task is
unblocked and can proceed independently of T02T04.
## Rename Production Inter-Hub Repo
```task
id: CORE-WP-0007-T02
status: wait
priority: high
state_hub_task_id: "16eb7ce3-b574-4667-a189-c14ff5d0502b"
```
Rename `inter-hub` to `inter-hub-haskell` after operator stabilization sign-off
and approval to retire the Inter-Hub rollback path. Compatibility, migration,
and production cutover gates from `CORE-WP-0005` are satisfied; the Inter-Hub
k8s deployment in namespace `inter-hub` still runs as rollback-only (ingress
already moved — see `k8s/railiance-production/README.md`).
Stabilization evidence is collected automatically through 2026-07-10 by the
scheduled `core_hub_stabilization_check` / `core_hub_stabilization_closeout`
jobs (see runbook `docs/deployment/stabilization-signoff.md`). T02 unblocks
when the closeout passes (`operator_signoff_needed=true`) and the operator
records retirement approval in State Hub.
Scope when unblocked: Forgejo repo rename, registry/Helm/CI path updates in
`railiance-apps` and `railiance-forge`, and eventual scale-down of the
rollback deployment after the stabilization window closes.
## Decommission Haskell Build Lane
```task
id: CORE-WP-0007-T03
status: wait
priority: medium
state_hub_task_id: "bea9943c-916f-402f-a6d9-7907bb48137c"
```
Retire or park haskelseed runner labels, the `haskell-build` VM profile,
IHP/GHC/Nix/devenv docs, and build-machine workplans after T02 completes and no
active repo still needs Haskell image builds. Primary touchpoints:
`railiance-forge` runner substrate docs, `sand-boxer` `profile.vm-haskell-build`,
and `ops-bridge` haskelseed tunnel notes.
## Update Cross-Repo References
```task
id: CORE-WP-0007-T04
status: wait
priority: medium
state_hub_task_id: "ed7f8e01-5305-465c-8bfc-04bc47bfdf40"
```
After T02, update docs and workplans in consuming repos so production Inter-Hub
references point to Core Hub and historical Haskell context explicitly names
`inter-hub-haskell`. Known stale surfaces: `ops-hub` INTENT/SCOPE/README and
readiness docs; `activity-core` railiance probes and sink vocabulary;
`the-custodian` service catalog and drain plans. Core Hub specs and migration
fixtures already use `inter-hub-haskell` as the bundle `source` name.
## Progress 2026-07-07 — upstream gate cleared
`CORE-WP-0005` archived (`260703-CORE-WP-0005-data-migration-and-cutover.md`).
Production cutover executed 2026-07-03: Core Hub serves `hub.coulomb.social`;
post-flip reconcile import was idempotent (0 created / 28 updated / 0 errors);
rollback remains one kubectl command while the Inter-Hub deployment stays
running through stabilization.
Revised sequencing:
1. **T01** — unblocked; archive `ihp-railiance-probe` first (low risk).
2. **T02** — cutover prerequisites met; blocked on operator stabilization
sign-off and approval to retire the Inter-Hub rollback deployment.
3. **T03** — blocked on T02 and confirmation that no repo still schedules
Haskell builds through haskelseed.
4. **T04** — blocked on T02 so references use the final repo name.
Related finished workplans: `CORE-WP-0004` (compatibility smokes),
`CORE-WP-0005` (migration/cutover), `CORE-WP-0008` (deployed ops evidence).
## Completion 2026-07-07 — T01 probe repo archived
`ihp-railiance-probe` retired in-repo:
- `README.md` marks the repo archived and points to preserved docs
- `INTENT.md` and `SCOPE.md` updated with retirement context
- `IRP-WP-0001` moved to
`workplans/archived/260707-IRP-WP-0001-pipeline-validation.md` (`status:
archived`)
- Gitea/Forgejo CI workflows switched to `workflow_dispatch` only (no push
triggers)
Forgejo repo archived on `gitea.coulomb.social` (2026-07-07); git history and
diagnostic docs remain available read-only.
## Progress 2026-07-07 — stabilization job live on railiance01
Post-cutover stabilization checks for T02 operator sign-off are implemented and
deployed. Runbook: `docs/deployment/stabilization-signoff.md`.
### Window and schedule
| Item | Value |
| --- | --- |
| Window start | 2026-07-03 (production cutover) |
| Window end | 2026-07-10T17:35:00Z |
| Daily check | 09:00 Europe/Berlin |
| Closeout | one-shot at window end |
Runs outside the window record `skipped` with `reason=outside_stabilization_window`.
### Automated checks (`core-hub` / `stabilization_check`)
| Probe | Expected |
| --- | --- |
| `GET /healthz` | 200 |
| `GET /readyz` | 200 |
| `GET /api/v2/widget-types` | 200, count ≥ 26 |
| `GET /api/v2/hubs` (unauthenticated) | 401 |
Closeout sets `operator_signoff_needed=true` when `overall_pass=true`.
### Activity definitions (canonical in core-hub)
| Job | ActivityDefinition id | State Hub `event_type` |
| --- | --- | --- |
| Daily | `b4c8e2f1-6a3d-4e5b-9f0c-2d7e8a1b3c4d` | `core_hub_stabilization_check` |
| Closeout | `c5d9f3a2-7b4e-5f6c-0a1d-3e8f9b2c4d5e` | `core_hub_stabilization_closeout` |
Files: `activity-definitions/core-hub-stabilization-{daily,closeout}.md`, projected
into `activity-core/k8s/railiance/20-runtime.yaml`. Progress events bind to
workstream `a8d66822…` / task `16eb7ce3…` (T02).
### Manual companion
```bash
make stabilization-check # scripts/core_hub_stabilization_check.sh
```
Optional: `make deployed-smoke` with `CORE_HUB_OPERATOR_TOKEN_FILE` for full
bootstrap evidence; `ops-hub` `make interhub-gate` against `hub.coulomb.social`.
### activity-core deployment (railiance01)
- Image: `activity-core:railiance01-prod` rebuilt and imported into k3s
- Resolver: `activity-core` `core-hub` context adapter (`stabilization_check`)
- Evidence sink fix: `core-hub` sources now post State Hub progress
(`activity-core` `726da05`)
- Deployed: `actcore-api`, `actcore-worker`, `actcore-event-router` rolled out;
`actcore-sync` registered both schedules on Temporal
- First live event (cluster State Hub
`http://state-hub.state-hub.svc.cluster.local:8000`, 2026-07-07T22:43:55Z):
`Core Hub stabilization daily: pass; base=https://hub.coulomb.social;
widget_types=26/26`
Note: activity-core posts to the **cluster** State Hub on railiance01, not the
workstation fallback at `127.0.0.1:8000`. Query via `state-hub-primary` tunnel or
from inside the `activity-core` pod when reviewing evidence.