[bitnami/kafka] How to define healthcheck for kafka container

[apachler]

Name and Version

bitnami/kafka:3.7

What is the problem this feature will solve?

I am using 3 instances of Kafka controlled by 3 instances of Zookeeper.

It seems the following healthcheck is not working if used in docker compose.

    healthcheck:
      test: ["CMD-SHELL", "kafka-topics.sh --list --bootstrap-server localhost:9092 || exit 1"]
      interval: 10s
      timeout: 30s
      retries: 10
      start_period: 30s

Since i am not very familiar with Kafka I just wanted to know what healthcheck command should be used.

What is the feature you are proposing to solve the problem?

Container image maintainers should provide a working healthcheck command.

[javsalgar]

Hi!

In the helm chart we have two checks:

liveness:

            exec:
              command:
                - pgrep
                - -f
                - kafka

readiness:

tcpSocket:
  port: client

So, for your case, you can try mimicking something similar in the container.

[apachler]

Is it really useful to check for a running process that is PID 1 ?
If PID 1 is not existing anymore the container would restart immediately. So this case is already catched.
Since i am not familiar with K8s i dont know what the readiness probe is doing exactly.

Something is really strange with Bitnami images because the initialization is not working/the container is not starting up if a HEALTHCHECK is enabled ...

[javsalgar]

In the case of the readiness I imagine that it does some sort of operation close to a telnet or similar.

[apachler]

Hmmm, I think that just opening a connection to Kafka is not enough to make sure that the service is correctly working.

Anyways, as already discussed in #75322 - something is really wrong with the Bitnami images because activating the Docker healthcheck is killing the whole functionality of the container...

[alvneiayu]

hi @apachler

let me check it and investigate a little bit, I will come back as soon as possible with some info.

Thanks

[igloo12]

I agree. My kafka pod is stuck where the readiness probe is failing, but the liveliness probe is okay.