For kafka bootstrap.servers config one should provide all the broker details.
If not, how does kafka producer/consumer will resolve all brokers available?
If yes, Do producer/consumer needs restart after adding a node?
1 Answers
4
votes
The bootstrap.servers config is mandatory for producer and consumer and it must have one broker host:port at least in order to allow producer/consumer to connect to the cluster and then getting all the other metadata (where topics are, who are the leaders and so on).
If you add a node, you don't need to restart producer/consumer but of course the current configuration will be stale. I mean ... If you have a cluster with broker1 and broker2 and you put them into the bootstrap.servers config then producer/consumer connect to the cluster. Let's imagine that you add broker3 and some new topics/partitions are deployed there. Your producer/consumer will be able to write/read to/from these new topics/partitions because information about them (so that they are on broker3) will be available through asking metadata to broker1 or broker2.
Of course if you shutdown broker1 and broker2 and your cluster will be just made of broker3, you'll have to restart producer/consumer with the bootstrap.servers config with just broker3.
The bootstrap.servers config is just (as the name implies) a configuration used on starting; then each client will be able to connect directly to a broker (for reading/writing topic/partitions there) even if such broker isn't in the bootstrap.servers config.
