---
title: "Comment by Rishabh Mishra on \"Stop Pinging Kafka - A Zero-Cost Health Check Pattern (and What Can Go Wrong)\" | Hashnode"
description: "One more guard worth adding: wire this check to the readiness probe rather than liveness. A false unhealthy then pulls the pod out of service instead of restart"
url: "https://hashnode.com/posts/stop-pinging-kafka-a-zero-cost-health-check-pattern-and-what-can-go-wrong/6ad65e65cb89e7ce1f6ae2c7/comment/6ad6d99c8b41f070ec2d91f0"
---

[Discussion](https://hashnode.com/posts/stop-pinging-kafka-a-zero-cost-health-check-pattern-and-what-can-go-wrong/6ad65e65cb89e7ce1f6ae2c7)

Comment [Stop Pinging Kafka - A Zero-Cost Health Check Pattern (and What Can Go Wrong)](https://hashnode.com/posts/stop-pinging-kafka-a-zero-cost-health-check-pattern-and-what-can-go-wrong/6ad65e65cb89e7ce1f6ae2c7)

[Rishabh Mishra](https://hashnode.com/@rushdeploy)

[Oct 20](https://hashnode.com/posts/stop-pinging-kafka-a-zero-cost-health-check-pattern-and-what-can-go-wrong/6ad65e65cb89e7ce1f6ae2c7/comment/6ad6d99c8b41f070ec2d91f0)

One more guard worth adding: wire this check to the readiness probe rather than liveness. A false unhealthy then pulls the pod out of service instead of restarting it, which avoids both the restart storm and the consumer rebalance the article describes.

[abhishek pareek](https://hashnode.com/@abhishek-pareek)

[Oct 20](https://hashnode.com/posts/stop-pinging-kafka-a-zero-cost-health-check-pattern-and-what-can-go-wrong/6ad65e65cb89e7ce1f6ae2c7/comment/6ad6d99c8b41f070ec2d91f0#reply-6ad6ee30d28d916a1e4b7502)

good point [Rishabh Mishra](https://hashnode.com/@rushdeploy "Rishabh Mishra") !

[Rishabh Mishra](https://hashnode.com/@rushdeploy)

[Oct 20](https://hashnode.com/posts/stop-pinging-kafka-a-zero-cost-health-check-pattern-and-what-can-go-wrong/6ad65e65cb89e7ce1f6ae2c7/comment/6ad6d99c8b41f070ec2d91f0#reply-6ad702f7c399e7d1392aa301)

Thanks! One detail worth pinning down: keep the liveness probe cheap, just a process-alive check. Put the Kafka state behind a readiness endpoint with a small consecutive-failure threshold, so one blip doesn't drain the pod. That pairing covers both failure modes without mixing them.
