Kafka Redpanda container not stopped and delete on application end

[ppatierno]

While developing an application using Kafka Streams using then Kafka Dev Services, we know that a Redpanda based Kafka broker is started in a container:

CONTAINER ID   IMAGE                         COMMAND                  CREATED        STATUS                      PORTS                                                   NAMES
634a0bfbe491   vectorized/redpanda:v21.5.5   "sh -c 'while [ ! -f…"   7 hours ago    Up 7 hours                  9644/tcp, 0.0.0.0:49154->9092/tcp, :::49154->9092/tcp   hungry_heisenberg

When is it supposed to be stopped and removed? When I stop my application it doesn't happen.
Is it related to the fact that I had to disable ryuk because of this other problem described here #22524 ?

[quarkus-bot[bot]]

/cc @cescoffier, @ozangunalp

[ppatierno]

I have docker on Fedora.

[ppatierno]

But as I mentioned on the other issue, who really likes to have a container running as a privileged container with rights of the current host? Tbh, I have never used a privileged container but AFAIU this is how ryuk runs.

[ppatierno]

I tried again this morning and even disabling ryuk (which doesn't work on my side, #22524) now the redpanda container is deleted when I stop the application. Tbh I don't know why, but it's cleaned even without ryuk. Is that expected somehow?

[cescoffier]

So, when we have a clean termination of the dev mode, we stop the container. However, there are cases where the clean-up cannot be executed (like a hard crash). In this case, ruyk is doing the cleanup.

[ppatierno]

Ok it makes more sense.

In this case, ruyk is doing the cleanup.

I am still not convinced that a dev wants a privileged container running to do the clean up, or at least not me. Anyway I still have the need to disable it otherwise my app doesn't work with dev services :-(

[ozangunalp]

Let me clarify two things here:

1/ Test containers does not run ryuk container in privileged mode by default. That is the reason it didn't work for you in the first place. Seems like there is a particular issue with Fedora: testcontainers/testcontainers-java#4014. I personally would qualify this as a non-issue as it doesn't impact most people and if it does there are multiple workarounds.

2/ There is actually two safeguards for stopping and deleting the dev service container. The first one is on the Quarkus end: we close and delete the container on the application exit. Second one is on the test containers end, with or without ryuk, it registers a runtime hook for cleaning up the container. Of course, if the process quits abruptly, (ex. with kill -9) the only thing that would work is ryuk which acts as a watchdog.

[ppatierno]

Thanks for the detailed explanation!