Show HN: Kafkatop 2.0 – top for Kafka – rewritten in Go with partition analytics

3 points

6 months ago

Kafkatop is a handy real-time monitoring CLI tool for Apache Kafka, that displays consumer lag and event rates in a clean, top-like interface. It can assess when consumers will catch up, and provides indicators for partition distribution health

v2 is a go rewrite from the original Python PEX. The main reason was the python version dependency on confluent's librdkafka which was not maintained any more for older systems. The new implementation is pure GO (franz-go, no CGO deps) meaning it's very safe to distribute.

The new features include:

* drill into partition details: config, lag, offsets, rates, replicas, ISR status, and leaders, with Cell colors indicating hotness

* partition health with two new metrics:

  - PAR (Peak-to-Average Ratio): shows if one partition is 5x, 10x, or 200x busier than average

  - Cv (Coefficient of Variation): measures overall topic skew from 0 (perfect) to >1 (critical)

* improved catch-up ETA calculation: accounts also for arrival rate

* better performance: parallel API calls with configurable concurrency (default 10 concurrent)

* custom DNS mappings: override DNS resolution for broker-provided kafka nodes allowing to work via port-forwarding.

The tool is here: https://github.com/sivann/kafkatop

Would love feedback, especially on the partition metrics - are PAR and Cv intuitive enough, or should I add different indicators? What other feature would you like to see?