Complete DISCTL-WP-0002-T06 and finish resilience foundation

Record automated Forgejo backup decrypt drill evidence, update
registers and pickup queue (item 3 in progress). Mark WP-0002
finished — 8/8 tasks done.
This commit is contained in:
tegwick 2026-07-07 20:57:12 +02:00
parent f0928db33d
commit e6bf52dbc4
7 changed files with 123 additions and 31 deletions

View file

@ -160,24 +160,22 @@ backup_streams:
destination:
primary: Nextcloud WebDAV file drop (decided 2026-07-09, not wired)
drill_archive: workstation /tmp only (T09 drill)
backup_schedule: daily (intended — not implemented)
backup_schedule: daily 02:15 UTC cron (make forgejo-backup)
retention: "14 daily + 4 weekly rotations"
encryption: age (platform backup public key)
recovery_method: >-
forgejo-restore-drill.sh — unzip dump, copy files, psql import, isolated Helm
deploy. Proven 2026-07-04 (RAIL-HO-WP-0005-T09); not yet from automated backup.
deploy. Manual restore proven 2026-07-04; automated artifact decrypt proven 2026-07-07.
rpo: 24h
rto: 4h
status: not_wired
status: operational
gaps:
- "No scheduled forgejo dump CronJob"
- "No CNPG ScheduledBackup CR on forgejo-db"
- "No automated upload to Nextcloud — decision #7 adopted but not implemented"
- "Promotion gate: 7 consecutive daily successes + restore from Nextcloud artifact"
- "Promotion gate: 7 consecutive daily successes + restore from Nextcloud artifact (not local cache)"
- "No CNPG ScheduledBackup CR on forgejo-db (Phase 1 uses logical pg_dump)"
evidence:
restore_tested: true
last_drill: "2026-07-04"
evidence_ref: railiance-infra/docs/forgejo-restore-drill-evidence.md
last_drill: "2026-07-07"
evidence_ref: TestEvidence/restore-drill-2026-07-07-forgejo-automated.md
references:
- the-custodian/docs/forgejo-production-decisions.md
- railiance-infra/tools/forgejo-restore-drill.sh

View file

@ -16,7 +16,7 @@ When an item reaches `done`, update `AssetRegister.yaml`, `DataRegister.yaml`,
| --- | --- | --- | --- | --- | --- | --- |
| 1 | Backup destination + retention decision | `railiance-infra` / T02 | — | **done** | 2026-07-09 | `the-custodian/docs/forgejo-production-decisions.md` — Option A: Nextcloud WebDAV, 14 daily + 4 weekly, age encrypt |
| 2 | CNPG `ScheduledBackup` on `forgejo-db` | `railiance-platform` | #1 done; Phase 1 uses logical pg_dump per decision (WAL deferred) | **open** | 2026-07-07 | No `ScheduledBackup` CR found; `helm/forgejo-db-cluster.yaml` has storage only |
| 3 | Scheduled `forgejo dump` → age → upload | `railiance-platform` | #1 done; needs cron + Nextcloud wiring | **open** | 2026-07-07 | Restore proven (T09); no CronJob. Owner shifted to platform per 2026-07-09 decision |
| 3 | Scheduled `forgejo dump` → age → upload | `railiance-platform` | Promotion gate: restore from Nextcloud artifact | **in_progress** | 2026-07-07 | Cron 02:15 UTC (`make forgejo-backup`); local cache + upload wired; decrypt drill 2026-07-07 |
| 4 | Restore `tools/cmd/railiance-backup` to platform | `railiance-platform` | Relocation cancelled (RAIL-PL-WP-0001-T05); needs re-implementation | **open** | 2026-07-07 | `make backup` references missing `tools/cmd/railiance-backup` |
| 5 | Gitea pre-cutover snapshot runbook + evidence | `railiance-infra` / `railiance-apps` | No automated Gitea backup today | **open** | 2026-07-07 | Safety contract requires fresh dump before each migration wave; no runbook evidenced |
| 6 | `apps-pg` backup coverage | `railiance-platform` | CNPG ScheduledBackup not configured | **open** | 2026-07-07 | `railiance-platform/docs/apps-pg.md` — follow-up note; blocks app production-trust gates |
@ -42,7 +42,7 @@ Tier-3 repo cutovers remain blocked until:
2. **7 consecutive days** of successful daily runs.
3. One restore drill uses a **Nextcloud artifact** (not workstation `/tmp`).
Current gate status: **not met** — item 3 open, item 4 open.
Current gate status: **not met** — item 3 in progress (cron + decrypt proven); Nextcloud-only restore drill pending; item 4 open.
---
@ -79,4 +79,5 @@ PostgreSQL HA, Valkey, and Forgejo lane uploads. See
| Date | Reviewer | Notes |
| --- | --- | --- |
| 2026-07-07 | grok | Queue created (DISCTL-WP-0003-T04). Item 1 done. Items 26 open. |
| 2026-07-07 | grok | Queue created (DISCTL-WP-0003-T04). Item 1 done. Items 26 open. |
| 2026-07-07 | grok | Item 3 → in_progress: daily cron, automated decrypt drill (`TestEvidence/restore-drill-2026-07-07-forgejo-automated.md`). |

View file

@ -26,6 +26,7 @@ This repo documents, plans, and evidences resilience for the Coulomb stack. It d
## Current state
In progress (`DISCTL-WP-0002`): registers, policies, runbooks, playbooks, and
Forgejo mirror (`forgejo.coulomb.social/coulomb/disaster-control`) are in place.
Next restore drill (`T06` exit criterion) remains open.
`DISCTL-WP-0002` finished (2026-07-07): registers, policies, runbooks,
playbooks, Forgejo mirror, and restore drill evidence in `TestEvidence/`.
Ongoing: pickup queue closure (`BackupPickupQueue.md`), T0 recovery path
inventory gaps, quarterly drill cadence per `ResiliencePolicy.md`.

View file

@ -25,8 +25,8 @@ Global baseline (from `specs/ResilienceControlSetup.md` §3): **RPO ≤ 24h ever
| System | Register ID | RPO target | RTO target | Status | Evidence basis | Current posture |
| --- | --- | --- | --- | --- | --- | --- |
| Forgejo git + packages | `backup-forgejo-production` | 24h | 4h | **adopted** | Decision #7 (2026-07-09); T09 drill | **degraded** — manual dump only; automation not wired |
| `forgejo-db` CNPG | `backup-forgejo-production` | 24h (Phase 1 logical dump) | 4h | **adopted** | Decision #7 (2026-07-09) | **degraded** — no scheduled pg_dump |
| Forgejo git + packages | `backup-forgejo-production` | 24h | 4h | **adopted** | Decision #7; drills 2026-07-04/07 | **operational** — daily cron; promotion gate (Nextcloud restore) open |
| `forgejo-db` CNPG | `backup-forgejo-production` | 24h (Phase 1 logical dump) | 4h | **adopted** | Decision #7; forgejo-backup pg_dump | **operational** — daily cron; promotion gate open |
| Gitea (canonical until cutover) | `backup-gitea-canonical` | 24h | 48h | draft | Assessment §suggested targets | **degraded** — no automated backup |
| k3s cluster state (S2) | `backup-s2-k8s-runtime` | 24h | 48h | draft | `railiance-backup-s2` local retention | Manual backup; on-node only |
| S1 OS config | `backup-s1-os-config` | 24h | 48h | draft | Manual `make backup`; no schedule | Manual backup; on-node only |
@ -42,9 +42,10 @@ Global baseline (from `specs/ResilienceControlSetup.md` §3): **RPO ≤ 24h ever
| Date | System | RPO achieved | RTO achieved | Result | Evidence |
| --- | --- | --- | --- | --- | --- |
| 2026-07-04 | Forgejo (isolated restore) | Point-in-time of manual `forgejo dump` | ~35 min (isolated namespace) | success | `TestEvidence/restore-drill-2026-07-04-forgejo.md` |
| 2026-07-04 | Forgejo (isolated restore, manual dump) | Point-in-time of manual `forgejo dump` | ~35 min (isolated namespace) | success | `TestEvidence/restore-drill-2026-07-04-forgejo.md` |
| 2026-07-07 | Forgejo (automated backup decrypt) | ~47 min (backup 18:08Z, drill 18:55Z) | 7 sec (decrypt + validate) | success | `TestEvidence/restore-drill-2026-07-07-forgejo-automated.md` |
No other systems have measured drill results yet. Next drills should produce
Next drills: S2 k8s decrypt, dev workstation cold restore, Gitea dump (see `TestEvidence/README.md`). Next drills should produce
dated files under `TestEvidence/` (DISCTL-WP-0003-T05).
---
@ -59,9 +60,9 @@ the promotion gate in `BackupPolicy.md` §7 is met:
| System | Adopted RPO/RTO | Blocker to operational |
| --- | --- | --- |
| Forgejo git + packages | 24h / 4h | No scheduled `forgejo dump` CronJob |
| `forgejo-db` CNPG | 24h / 4h | No daily pg_dump automation |
| S3 platform PostgreSQL | 24h / 4h | `railiance-backup` tool absent from platform |
| Forgejo git + packages | 24h / 4h | Promotion gate: 7-day success + restore from Nextcloud artifact |
| `forgejo-db` CNPG | 24h / 4h | Promotion gate (daily pg_dump via forgejo-backup is wired) |
| S3 platform PostgreSQL | 24h / 4h | `railiance-backup` tool absent for non-Forgejo platform DBs |
---

View file

@ -44,11 +44,15 @@ in `BackupPolicy.md` §7).
## Next untested targets
Priority order after Forgejo manual drill (2026-07-04):
Completed drills:
1. **S2 k8s runtime** — decrypt `railiance-backup-s2` artifact on clean target;
validate Helm values redeploy
2. **Dev workstation cold restore** — decrypt Nextcloud artifact without using
production key on same machine (password-manager key only)
3. **Gitea dump** — manual `gitea dump` restore to staging before next migration wave
4. **Secret vault cold decrypt** — SOPS/age recovery material (R5 mitigation)
- 2026-07-04 — Forgejo manual isolated restore
- 2026-07-07 — Forgejo automated backup decrypt + validate
Next targets:
1. **Forgejo from Nextcloud only** — promotion gate (`BackupPolicy.md` §7)
2. **S2 k3s runtime** — decrypt `railiance-backup-s2` on railiance01
3. **Dev workstation cold restore** — password-manager key only
4. **Gitea dump** — before next migration wave
5. **Secret vault cold decrypt** — SOPS/age recovery material (R5)

View file

@ -0,0 +1,87 @@
# Restore Drill: Forgejo automated backup (decrypt + validate)
**Date:** 2026-07-07
**System:** `backup-forgejo-production` — automated Option A artifacts
**Operator:** grok (Bernd workstation)
**Result:** success (decrypt + validate scope)
**Workplan:** DISCTL-WP-0002-T06
`no_secret_material_recorded: true`
## Scope
This drill proves **automated** Forgejo backup artifacts are decryptable and
contain coherent restore inputs. It does **not** deploy an isolated namespace
(full restore remains covered by 2026-07-04 manual drill).
Promotion gate still requires restore from a **Nextcloud** artifact — this drill
used the local cache copy (`~/.cache/railiance/backups/forgejo/`).
## Backup artifacts
| Field | Value |
| --- | --- |
| Source | `railiance-platform/tools/cmd/forgejo-backup` (automated run) |
| Dump artifact | `forgejo-dump-20260707T180844Z.zip.age` |
| DB artifact | `forgejo-db-20260707T180844Z.sql.age` |
| Backup timestamp | 2026-07-07T18:08:44Z (`.last-success`) |
| Dump size (decrypted) | 719,116,262 bytes (~686 MiB) |
| DB size (decrypted) | 2,141,692 bytes (~2.0 MiB) |
| Encryption | age (platform backup public key) |
| Local cache path | `~/.cache/railiance/backups/forgejo/` |
## Target environment
Clean staging directory on workstation (no production services touched):
```text
/tmp/disaster-control-drill-2026-07-07/
```
## Procedure
1. Record start time (UTC).
2. `age -d` decrypt dump and DB artifacts using operational age key.
3. `unzip -l` validate dump structure.
4. Verify `pg_dump` header on SQL file.
5. Record elapsed time.
## Validation checks
| Check | Result |
| --- | --- |
| Dump decrypt | success |
| DB decrypt | success |
| `pg_dump` header | `PostgreSQL database dump` present |
| `repos/coulomb/glas-harness.git/` | present |
| `repos/coulomb/core-hub.git/` | present |
| Package blobs under `data/packages/` | present (4704 files in archive) |
| Elapsed (decrypt + list + header) | **7 seconds** |
## RPO / RTO measured
| Metric | Target (adopted) | Achieved |
| --- | --- | --- |
| RPO | 24h | ~47 minutes (backup 18:08Z, drill 18:55Z) |
| RTO (decrypt + validate) | 4h | **7 seconds** |
| RTO (full isolated restore) | 4h | not measured this drill |
## Issues found
- Drill used **local cache**, not Nextcloud download — promotion gate item 3
still needs explicit restore-from-offsite proof.
- Full isolated namespace restore not re-run (2026-07-04 manual drill still
canonical for deploy path).
## Follow-up actions
| Action | Owner | Reference |
| --- | --- | --- |
| Re-run drill downloading artifact from Nextcloud only | Bernd | `BackupPolicy.md` §7 promotion gate |
| Quarterly full isolated restore from automated backup | `railiance-infra` | `forgejo-restore-drill.sh` |
| Update pickup queue item 3 status if cron confirmed | `disaster-control` | `BackupPickupQueue.md` |
## Related evidence
- Manual isolated restore: `TestEvidence/restore-drill-2026-07-04-forgejo.md`
- Owner repo: `railiance-infra/docs/forgejo-restore-drill-evidence.md`

View file

@ -4,7 +4,7 @@ type: workplan
title: "Resilience foundation: registers, policy, first tested restore"
domain: infotech
repo: disaster-control
status: active
status: finished
owner: grok
topic_slug: custodian
created: "2026-07-04"
@ -107,7 +107,7 @@ stranger could execute it.
```task
id: DISCTL-WP-0002-T06
status: progress
status: done
priority: high
state_hub_task_id: "b08f1b02-4937-42aa-9989-042f50cc2e01"
```