Incident Lab · SEV2 · Databricks
GMV down 62% today
Finance: today's GMV tile dropped 62% versus yesterday. Is something wrong with sales or with the data?
Job daily_gmv_refresh started 2026-08-03 06:00 while ingest_sales_events (ds=2026-08-03) is RUNNING, 2 of ~5 files committed
daily gmv refreshes at 06:00, but the ingest job for today's partition of sales events hadn't finished. The model aggregates whatever files are there, so today shows a fraction of the real number. The ingest job writes its progress to load audit : one row per attempt, status is running , failed or complete . (Spark jobs often signal the same thing with a SUCCESS file.) Investigate: Fix daily gmv so it only reports partitions that have a complete load. Columns: ds , orders (row count), gmv (sum of amount , rounded to 2), ordered by ds . Each day must still count once, even if it took more than one attempt.