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
- Fix score_pressure crash on nil pressure_mb (coastal HRRR points)
- Set 10-min timeout on grid score upsert transaction (was :infinity)
- Single DELETE for prune_old_scores instead of N queries in a loop
- Remove dead load_hrrr_refractivity that loaded 95k rows into nil map
- Pass selected_time to point_detail to skip latest_valid_time sub-query
- Batch station existence checks (1 query per path point, not per station)
- Batch solar index upserts via insert_all in chunks of 500
- Batch backfill_distances via single UPDATE FROM VALUES statement
- Add is_grid_point boolean + partial index to hrrr_profiles (replaces
non-sargable modular arithmetic filter on every weather map query)
- Add partial index on contacts(qso_timestamp) WHERE pos1 IS NOT NULL
- Move backfill enqueue to Oban worker so UI returns immediately
Removed n_live_tup > 0 filter that hid tables when pg_stats reports 0 rows
during VACUUM. Added dead tuple column with warning highlight to monitor
vacuum progress. Also includes Styler reformatting and unused dep cleanup.
ON CONFLICT now only replaces score/factors when the score actually changed,
avoiding dead tuple generation for the ~80% of grid points that don't change
between consecutive HRRR runs.
Migration sets aggressive autovacuum on propagation_scores: zero cost delay,
2000 cost limit, 1% scale factor. The table was at 72GB with 108M dead rows
because default autovacuum couldn't keep pace with 14M upserts per hour.
