Consider two calls to provision running in parallel. Both detect a topic doesn't yet exist in the cluster. Both try to create it. The first to try succeeds, the second will fail with a 'topic already exists' error.
Currently, this error is propagated to the caller. A better solution would be to gracefully handling this issue and just checking the topic has been created with the right config.
Consider two calls to provision running in parallel. Both detect a topic doesn't yet exist in the cluster. Both try to create it. The first to try succeeds, the second will fail with a 'topic already exists' error.
Currently, this error is propagated to the caller. A better solution would be to gracefully handling this issue and just checking the topic has been created with the right config.