Grokking Kafka and Event-Driven Architecture
Vote

0% completed

​

Notification System: Failures and Operations

At 18:00 on a busy evening, the SMS provider's typical call slows from 400 ms to 4 seconds, and it stays slow for an hour. What should happen to the fraud checks waiting in sms-urgent, to the transaction alerts in sms-alerts, and to the engineer who is on call?

Two answers are tempting and both are wrong. Doing nothing lets a backlog grow at 120 alerts a second. Retrying harder, or adding many more consumers, may push an overloaded provider further behind. The right answer depends on numbers the design already has: the call time, the consumer count, and each message's expiry.

.....

.....

.....

Like the course? Get enrolled and start learning!

Reading Progress

0%

Vote for new content