Outage replay of a stadium-tour on-sale

Views in this map

1. 1
   18:55 NormalFive minutes to the on-sale: 4,000 requests a second, every service healthy.
2. 2
   19:00 Sale opens1.2 million fans arrive at once. Traffic jumps 40× and the seat path starts to strain.
3. 3
   19:04 Inventory DB saturatesSeat-lock writes pile up on hot rows; holds time out and checkouts fail.
4. 4
   19:11 Waiting room onThe waiting room lets fans in at a set rate; the database recovers and sales resume.
5. 5
   19:40 RecoveredEvery service is healthy again, and the sale runs on through the waiting room.

18:55 ‹ › 19:00 ‹ › 19:04 ‹ › 19:11 ‹ › 19:40

Flows to play

• ▷
  Buy two ticketsOne fan buys two seats in section 112, from the seat map to the tickets in their wallet. In every view, so the same purchase goes fine, slows, fails at the seat lock, waits in line and goes fine again.

The model behind it

52 records of 13 types. Every box on the map is one of them, so the filters and the record panel read the same data.

• Services 11
• Teams 7
• Timeline moments 7
• Data stores 5
• Alerts 5
• Third parties 4
• Follow-ups 4
• 6 more types 9

Filters

Team and Tier: pick a value and everything else fades.