Backend & System Design Preparation

Marketing launched a heavily promoted 50%-off sitewide sale 10 minutes ago. However, the checkout service is inexplicably still charging users the full original price. Customers are abandoning their carts in anger, threatening a massive PR disaster on social media.

We recently deployed a new aggressive caching layer at the edge to improve profile load times. Unfortunately, users are now reporting a terrifying privacy issue: when they log in to their dashboard, they occasionally see names, settings, and avatars of completely different users.

A white-hat security researcher just disclosed a critical vulnerability in our Single Sign-On flow. By manipulating the OAuth callback URL, they demonstrated how an attacker can link their own external social login provider to any victim's account, allowing a complete account takeover.

It is Diwali Night, the biggest event of the year. At 7:00 PM, traffic spiked to 3.5x normal levels as thousands of users tried to place dinner orders simultaneously. The database CPU hits 100% and connection pools are exhausted, causing a catastrophic cascading failure.

A mega-influencer just joined the platform and asked their external audience to follow them. They gained 500k followers in an hour. However, the profile page shows erratic follower numbers—sometimes 500k, sometimes 12k, and sometimes 450k, depending entirely on which cache node serves the request.

You just launched a massive multi-million dollar ad campaign during a live sporting event. Traffic spiked 50x in seconds as users flooded the app. Immediately, the authentication service crashed under the weight of password hashing, locking everyone out.

It is payday for the platforms creators. The weekly giant payout job ran overnight and reported a glowing success. However, finance just realized that 10,000 creators never received their earned money. Worse, absolutely zero alarms fired, and there are no logs indicating why they were skipped.

A scheduled background cron job responsible for capturing monthly subscription payments failed mid-execution yesterday. When operations blindly restarted the job today, thousands of users were charged a second time, triggering a wave of furious customer support tickets.

Over the past month, the home feed has been degrading. Today, the P99 latency for the main feed endpoint hit a staggering 10 seconds. Users are staring at infinite loading spinners, and engagement metrics are plummeting. The database read replica IOPS are constantly maxed out.

A new viral challenge is sweeping the platform. Millions of users are simultaneously trying to upload their massive 4K video clips. The background processing queue has backed up by 12 hours, and the main upload API is completely unresponsive, throwing 503 errors to all clients.

A malicious user account has somehow bypassed frontend limits and added 50,000 unique items to their shopping cart. Every time this user logs in or attempts to view their cart, the entire pricing service runs out of memory and crashes, impacting thousands of regular shoppers.

The annual Midnight Madness flash sale went live 15 minutes ago. Traffic surged 10x instantly. While the servers are holding up, the finance team just flagged a critical data integrity issue: Customers are being charged multiple times for the same order due to race conditions.