usr/local/share/kafka/plugins directory on each machine runningĬonnect and then place the plugin directories (or uber JARs) there. Using the plugin path example above, you would create a Or, you can update the plugin path by adding the absolute path of the directoryĬontaining the plugin. That resolves to one of these) in a directory already listed in the plugin path. To install a plugin, place the plugin directory or uber JAR (or a symbolic link Plugin.path=/usr/local/share/kafka/plugins Is safely replicated, losing the node where a Connect worker runs does not And, because Kafka Connect stores connectorĬonfigurations, status, and offset information inside the Kafka cluster where it If a node unexpectedly leaves theĬluster, Kafka Connect automatically distributes the work of that node to You can add more nodes or remove nodes as your needs evolve.ĭistributed mode is also more fault tolerant. Kafka Connect distributes running connectors across the cluster. It can also be used for environments that typically use single agents (for example, sending web server logs to Kafka).ĭistributed mode runs Connect workers on multiple machines (nodes). Standalone mode is useful for development and testing Kafka Connect on a local machine.
Which mode works best for your environment before getting started. Running workers: standalone mode and distributed mode. Process is called a worker in Kafka Connect. Connectors and tasks are logical units of work and run as a process.
