Event Bus
System Analysis
Normal Behavior
An intermediary module that translates, routes, and queues messages between different applications or services.
Failure Behavior
Quietly drops a million critical financial transactions into a dead letter queue and forgets to tell anyone.
Business Consequence
Orders are marked as paid but never shipped because the fulfillment service never received the event.
Visual Manifestation
"A rapidly overflowing queue graph that suddenly drops to zero without any corresponding success metrics."
Satirical Behavior
"The magical black box where developers dump all their data and say 'it's asynchronous now', completely ignoring the fact that when it inevitably fills up and crashes, every single microservice will burn to the ground simultaneously."
Known Aliases
Technical Terminology
Failure Indicators
System Architecture (Graph)
Used By (Characters)
FAQ
How does it normally behave?
An intermediary module that translates, routes, and queues messages between different applications or services.
How does it fail?
Quietly drops a million critical financial transactions into a dead letter queue and forgets to tell anyone.
What is the business consequence?
Orders are marked as paid but never shipped because the fulfillment service never received the event.
What causes consumer group rebalance storms in partition-based brokers like Apache Kafka and how do you resolve them?
A rebalance storm occurs when a consumer takes longer to process a batch of records than the configured max.poll.interval.ms. The group coordinator assumes the consumer died, evicts it, and triggers a group-wide partition rebalance, pausing message consumption across all consumers. As queues back up, other consumers also exceed timeouts, causing an infinite rebalance cycle. This is resolved by increasing max.poll.interval.ms, reducing batch sizes (max.poll.records), and ensuring asynchronous message processing.
What is the Dead Letter Queue (DLQ) pattern and why is it essential for message broker reliability?
A Dead Letter Queue is a dedicated secondary queue where messages that fail processing repeatedly (e.g., due to corrupt schemas or bad payloads known as poison pills) are automatically routed after exhausting a retry threshold. Without a DLQ, a poison pill message causes consumers to crash repeatedly, blocking the entire partition and preventing healthy subsequent messages from being processed.
Explore the system
AI Summary
Message Broker / Event Bus is a MESSAGING system in TinyCTO.tv. Producers publish structured messages or events to topics. The broker persists messages to durable storage across partitioned logs, coordinates consumer offsets, and delivers messages to consumer groups with configurable delivery guarantees (at-least-once, at-most-once, exactly-once semantics).