Hello, Considering the following setup: 3x zk nodes running 3.5.0-alpha (for the ability to reconfig without shutting down)2x kafka nodes 0.8.2.1 each topic has 10 partitions, replication factor =2 Deploying on GKE, so to automate broker ID definition, I used the IP without the dots as a unique broker.id Now, when a broker dies, it gets re-started, and its IP changes.In my understanding, with a replication factor of 2, if a broker dies, when a new one joins the kafka cluster, the data gets replicated to it automatically, is that correct? What I see in zookeeper though, is that my topics still seem to reference the original broker.id, which died. If broker disappear, shouldn't kafka update the reference to the topic broker in zk? More troublesome, when both kafka broker are shut down, and restarted, now I have a new Kafka cluster, and the data is not replicated, but topics created with the old ids are still in zookeeper, and if I consume from an old topic, kafka can't find the broker. So the questions is:- should one broker fail and be replaced, does kafka automatically update the topic leader/follower broker ids, or is this a 'manual' operation I should take care of? - should both broker fail (then restarted with different ids, but still able to mount the original data volume, therefore they have the old data) is there a 'manual' way to update the topic broker references?
Thanks.
