Uses Repo, Scorer, BandConfig, Weather directly instead of raw SQL.
Scores all historical contacts with the real composite_score/2 and
includes factor heatmaps and weight sensitivity analysis.
Connects to local dev DB to analyze contacts, HRRR conditions,
terrain impact, commercial link correlation, and score sensitivity.
Ready for use once prod data is imported locally.
Apply 5-point circular moving average to reach_km values before
rendering the boundary polygon. Eliminates sharp spikes caused by
isolated terrain anomalies in SRTM data.
Runs BackfillEnqueueWorker every 30 minutes to pick up contacts with
pending/queued/failed enrichment status and enqueue missing jobs.
Before enqueueing, reconciles terrain contacts stuck in "queued" by
checking the terrain_profiles FK directly.
When ASOS/RAOB/IEMRE APIs return empty data, store a stub record so
dedup checks (has_surface_observations?, has_sounding?, has_iemre_observation?)
see coverage and don't recreate the same jobs on future backfill runs.
Queued contacts just need reconciliation (check if data exists),
while pending contacts need actual HRRR fetches. Process pending
first so real work happens before reconciliation.
Backfill dashboard now has checkboxes to select which enrichment types
to process (HRRR, Weather, Terrain, IEMRE). Prevents weather jobs from
flooding the queue when only HRRR needs work.
The propagation grid stores profiles at 0.125° intervals, so contacts
up to 0.0625° from a grid point would miss with the old 0.05° tolerance.
This caused has_hrrr_profile? and find_nearest_hrrr to fail matching
against existing data, leaving contacts incorrectly stuck as pending.
enqueue_for_contact unconditionally marked contacts as :queued even
when no HRRR jobs were needed (profiles already exist). Workers never
updated contact status to :complete after storing profiles. Contacts
got stuck and were never re-processed.
- Mark :complete when no jobs generated (data already exists)
- Include :queued in backfill enrichable states for reconciliation
Per-pod queues divided by 3 to maintain sane cluster totals.
Lifeline rescue reduced to 10 min to recover OOM orphans faster.
Memory limit 2Gi → 4Gi for HRRR wgrib2 processing.
- Add ecto_psql_extras for PostgreSQL insights in dashboard
- Add os_mon for OS metrics (CPU, memory, disk)
- Fix progress bars using percentage instead of raw large numbers
maxSurge:1 + maxUnavailable:0 deadlocks with anti-affinity since
there's no free physical host for the surge pod. Switch to
maxSurge:0 + maxUnavailable:1 to terminate one old pod first.
- libcluster with Kubernetes IP strategy for pod discovery
- RBAC for pod list/get, POD_IP from downward API, shared cookie
- LiveDashboard at /dashboard for all environments
