Allegro/2018-08-01/hours
A marketing campaign DDoSed itself
Jan 2018Dec 2018
Poland's largest e-commerce site went down at the moment a campaign was driving peak traffic.
Lesson
Capacity you can't actually schedule isn't capacity. Test that autoscaling can really claim idle resources under real load.
Trigger
A marketing push caused a sudden traffic spike.
Mechanism
A misconfiguration in cluster resource management stopped new service instances from starting — so even with free hardware available, the platform couldn't scale to meet demand and fell over.
Interview lens
Capacity only helps if the scheduler can use it. Test that new instances can start during expected traffic spikes.
Recurring patterns
