#57 · Why did your iPhone ask for the password? The Apple outage (17 Nov 2026)
iCloud sign-in, App Store and Apple Music went down together, and Apple blamed an internal update. Why one login service can take everything down, and the staged-rollout answer interviewers want.
Ek line yaad rakho: darwaza ek ho, toh chaabi sambhal ke badlo.
What happened
- 17 Nov 2026, from about 11:45 a.m. ET (around 10:15 PM IST): Apple's status page listed iCloud Account & Sign In as affecting all users. The App Store, Apple Music, AppleCare, iCloud Mail and iCloud.com were also hit; people were signed out or asked for their Apple Account password, and some couldn't activate new iPhones (TechCrunch). Purchases in Apple's retail stores failed too (MacRumors).
- Cause, per Apple: "an internal update caused some outages", not a cyberattack. Everything showed as resolved by about 6 PM ET.
Why one update hits everything
Every service first asks "who is this user?", and one shared login service answers. When it fails, the services behind it are healthy but useless. Apple hasn't published its internals; this is the classic interview reading of the symptoms.
The interview answer: single point of failure + staged rollout
- Single point of failure: one shared dependency that everything needs. Find these first in any design.
- Staged rollout: ship a change to ~1% of users or one region, watch error rates, then widen. Roll back automatically when errors jump.
- Graceful degradation: don't log people out when the login service is sick; let existing sessions keep working on their current token.
- Blast radius: split the service into independent cells so a bad update hits one cell, not everyone.
Words to say: single point of failure, staged rollout, canary, rollback, graceful degradation, blast radius.
Follow-up questions
You can't remove the login service. How do you make it safer? Run it in independent cells (by region or user shard), deploy to one cell at a time, keep tokens verifiable locally (signed tokens) so other services don't call it on every request, and cache recent results.
Why not just make every service do its own login? Then you have many copies of security-critical code to keep in sync. Centralised identity is right; it just needs the strictest rollout process in the company.
Illustrations simplified; not affiliated with Apple.
Go deeper
Part of the Concepts track (How systems and AI systems work).
Finished reading?
Mark it done to track your progress in Reels.
#59 · Six days, seven outages: not hacks, changes
Microsoft, the FAA, United, Chase, Apple, Marriott and the NSA all had outages between 12 and 17 Nov 2026. Almost every stated cause was a change. Safe deployment, rollback and blast radius.
#51 · Thank you Microsoft — the Microsoft 365 outage (12–13 Nov 2026)
Outlook, Teams calendars and Copilot down for 18+ hours. Servers failing health checks, traffic piling onto the rest, and the cascading failure pattern every backend interview asks about.