Kafka does not start correct

[fritterhoff]

In some cases the kafka container does not startup correctly so the services are unable to publish/... their data.

An indicator for that situation seems to be the fact that kafka starts a new broker with id != 1001 on the inital start and the services run into a timeout.

At the moment the following workaround seems to fix the problem:

• Stopping kafka
• Getting the implicit created docker volume by running docker volume kafka and note the volume id
• Executing a docker-compose down kafka
• Deleting the volume by the id

At the moment this issue seems to affect Windows 10 + Docker and Debian 10 + Docker

[DerAlexmeister]

What steps do you took to get to this point?

[fritterhoff]

The Issue is not reproducible with 100% but it seems like that stopping the kafka container fails sometimes. On the next startup some "fragments" from the previous execution obviously cause the issues.

[fritterhoff]

Ah okay. Seems like, I figured out whats the problem. If you change anything on the kafka container, docker-compose recreates the container. That causes creating a new broker and this broker seems to be useless.

So after changing anything on the docker-compose configuration of kafka you must delete the old container including the volume.

So the steps to reproduce:

1. Start kafka
2. Stop kafka
3. Change config (e.g. IP,...)
4. Start kafka again
5. Kafka will propably be broken even if you revert the changes in step 3

[DerAlexmeister]

Workaround:

1. Stop all container
2. Run sudo docker system prune (Only Linux/Unix) :)
3. Restart Zookeeper
4. Restart Kafka
5. Restart YAFRA
6. Working?