tracksolid_timescale_grafan.../260427_device_reconciliation.md

92 lines
3.3 KiB
Markdown
Raw Normal View History

feat(analytics): Phase 0 — analytics-config migration and CSV importer rewrite Phase 0 of the three-stakeholder analytics redesign: - 08_analytics_config.sql: ops.cost_rates + ops.kpi_targets with seed fuel rates (KES 195/L NBO+MBA, UGX 5200/L KLA) and 6 seed KPI targets (utilisation_pct, idle_pct global+osp-patrol, fuel_kes_per_100km, mttr_hours, alarms_per_100km). Granted SELECT to grafana_ro. Wired into run_migrations.py MIGRATIONS. - import_drivers_csv.py: full rewrite for the new Mitieng CSV (20260427_FSG_Vehicles_mitieng.csv). Snake_case columns, drops _infer_city() plate-prefix logic in favour of reading assigned_city directly. Adds cost_centre, assigned_route, vehicle_category, vehicle_brand, fuel_100km, depot_address. Treats the literal "NULL" string as missing. Reuses clean(), clean_num(), clean_ts(), get_conn(), get_logger() from ts_shared_rev. Special-cases numeric and timestamptz columns in the UPDATE clause. - audit_device_reconciliation.py: read-only audit comparing the CSV against tracksolid.devices. Reports per-account row counts, IMEIs on one side only, and devices on both sides whose metadata is still NULL. - 260427_device_reconciliation.md + 260427_audit_output.txt: Phase 0.2 reconciliation record. First run: DB has 172 devices, CSV has 162, delta +10 (10 IMEIs in DB-only, mostly fireside-account auto-syncs). Importer run with --only-null --apply filled 154 rows; coverage now assigned_city 152/172, cost_centre 150/172. Applied to stage on 2026-04-27 23:35 UTC. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-04-27 20:42:37 +00:00
# Device Reconciliation — 162 vs 182 (2026-04-27)
Phase 0.2 of the Business Analytics redesign. Resolves the gap between
`20260427_FSG_Vehicles_mitieng.csv` (162 rows) and `tracksolid.devices`
(~182 rows at last check).
## How to populate this report
1. Pull this branch onto the Coolify host (or rebuild containers so the
ingest container has `audit_device_reconciliation.py`).
2. Run inside the ingest container so it has `DATABASE_URL`:
```bash
ING=$(docker ps --filter name=ingest_movement --format "{{.Names}}" | head -1)
docker cp 20260427_FSG_Vehicles_mitieng.csv "$ING":/app/
docker exec "$ING" python audit_device_reconciliation.py \
--csv 20260427_FSG_Vehicles_mitieng.csv \
--out /tmp/260427_audit_output.txt
docker cp "$ING":/tmp/260427_audit_output.txt ./
```
3. Paste the audit output into the **Audit output** section below.
4. Mark the chosen disposition for each IMEI in the **Disposition** section.
## Audit output
First run: 2026-04-27 23:35 UTC against `tracksolid_db` on stage. Full output in
`260427_audit_output.txt`. Headline numbers:
| Metric | Value |
|---|---|
| CSV rows | 162 |
| `tracksolid.devices` rows | 172 |
| Delta (DB CSV) | +10 |
| In CSV but not in DB | 0 |
| In DB but not in CSV | 10 |
| Devices both sides, DB metadata still NULL on ≥1 field | 162 (resolved by importer run) |
After running `import_drivers_csv.py --only-null --apply` (2026-04-27): 154
devices updated, 8 already complete, 0 inserted. Coverage now: `assigned_city`
152/172, `cost_centre` 150/172, `vehicle_brand` 2/172, `fuel_100km` 3/172.
`assigned_route` / `vehicle_category` / `depot_address` remain 0/172 (CSV
provided no values for these — Phase 1 follow-up).
The 10 in-DB-not-in-CSV IMEIs are listed in `260427_audit_output.txt`. They
sit in `(blank)` or `fireside` accounts and surface in Grafana as
`assigned_city = 'unassigned'` thanks to the existing COALESCE in
`07_analytics_views.sql`.
## Disposition
For each IMEI in the "in DB but not in CSV" list, choose one and record why:
| IMEI | Account | Last seen | Disposition | Notes |
|---|---|---|---|---|
| | | | | |
**Disposition options:**
- **prune** — Delete from `tracksolid.devices`. Use when the unit is a stale
test/decommissioned device that should never have synced. Capture the SQL
before running:
```sql
DELETE FROM tracksolid.devices WHERE imei = '<imei>';
```
*Caveat:* foreign keys from `position_history`, `trips`, `alarms` must be
considered first — these will block the delete if there's any history.
Usually safer to leave-as-NULL.
- **leave-as-NULL** — Keep the row; metadata fields stay NULL. The device
was auto-synced from a Tracksolid account that the CSV doesn't cover
(likely `Fireside@HQ` rows that were left out of this Mitieng export).
Grafana views already use `COALESCE(d.assigned_city, d.city, 'unassigned')`
so these surface as "unassigned" but don't break dashboards.
- **addendum** — Add to a follow-up CSV and re-run the importer with
`--apply`. Use when the device is legitimate fleet metadata was just
missing from the export.
## Decision
<!-- Fill in once dispositions are chosen. -->
- Total devices reviewed: ___
- Pruned: ___
- Left-as-NULL: ___
- Added via addendum: ___
After action, re-run `audit_device_reconciliation.py` and confirm the
delta is what you expect.