# Kafka

> Apache Kafka Skill

- Skill: `ashish7802/kafka` (Agent Skill, multi-file: 9 files)
- Install (CLI): `npx skillmds@latest add ashish7802/kafka`
- Raw SKILL.md: https://api.skillmd.com/api/skills/ashish7802/kafka/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: ashish7802 (https://skillmd.com/u/ashish7802)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/ashish7802/kafka

---

# Apache Kafka Skill

> Distributed event streaming platform.

## Ecosystem Graph Preview

```mermaid
graph LR
  kafka["kafka"]:::core
  classDef core fill:#f9f,stroke:#333,stroke-width:4px;
  kafka -- "alternative to" --> rabbitmq
  kafka -- "alternative to" --> redis-streams
  nats -- "related to" --> kafka
  rabbitmq -- "alternative to" --> kafka
  redis-streams -- "alternative to" --> kafka
```

## Recommended Next Skills

- **[rabbitmq](/skills/rabbitmq)** (Score: 0.9)
  *Why: Direct relationship, Both are Messaging, Shared ecosystem (infrastructure), Can deploy to aws*
- **[redis-streams](/skills/redis-streams)** (Score: 0.8)
  *Why: Direct relationship, Both are Messaging, Can deploy to aws*
- **[nats](/skills/nats)** (Score: 0.7)
  *Why: Direct relationship, Both are Messaging*

## Quick Start
Kafka is not a queue; it is a distributed, append-only log. It excels at handling massive event streams (like clickstreams or IoT telemetry) by distributing partitions across a cluster.

```bash
docker run -p 9092:9092 apache/kafka:latest
```

## Production Patterns
### Consumer Groups and Partitions
Kafka guarantees message ordering *only within a single partition*. To scale out, you must increase the number of partitions for a topic. The maximum number of concurrent consumers in a group is strictly equal to the number of partitions.

## Architecture & Scaling
### Log Retention
Unlike RabbitMQ where messages are deleted once consumed, Kafka retains messages on disk based on a retention policy (e.g., 7 days or 100GB). This allows new consumer groups to replay the entire history of events from the beginning.

## Error Recovery
If a consumer is too slow, Kafka will trigger a 'rebalance', pausing message processing for the entire group. Ensure your consumer's `max.poll.interval.ms` is tuned properly, and keep processing logic completely asynchronous from the polling loop.

## Security Notes
Enable SASL/SCRAM for client authentication and TLS for in-transit encryption. Kafka without security is completely open to the network.

## References
- [Kafka Docs](https://kafka.apache.org/documentation/)

## Why use this skill
Use this when your agent works with **kafka** — structured patterns beat pasted docs and prevent common hallucinations.

## AI pitfalls
- Using outdated SDK or API versions from training data
- Inventing environment variable names
- Omitting error handling and retry logic

## Production checklist
- [ ] Secrets in environment variables, not source code
- [ ] Error handling and logging in place
- [ ] Rate limits and timeouts configured

## Related skills
- [`rabbitmq`](../rabbitmq/SKILL.md) — alternative to
- [`redis-streams`](../redis-streams/SKILL.md) — alternative to

---
> **Last Verified:** 2026-07-02