What Are the Differences Between the Client Mode and Cluster Mode of Spark Jobs?¶
You need to understand the concept ApplicationMaster before understanding the essential differences between Yarn-client and Yarn-cluster.
In Yarn, each application instance has an ApplicationMaster process, which is the first container started by the application. It interacts with ResourceManager and requests resources. After obtaining resources, it instructs NodeManager to start containers. The essential difference between the Yarn-cluster and Yarn-client modes lies in the ApplicationMaster process.
In Yarn-cluster mode, Driver runs in ApplicationMaster, which requests resources from Yarn and monitors the running status of a job. After a user submits a job, the client can be stopped and the job continues running on Yarn. Therefore, the Yarn-cluster mode is not suitable for running interactive jobs.
In Yarn-client mode, ApplicationMaster requests only Executor from Yarn. The client communicates with the requested containers to schedule tasks. Therefore, the client cannot be stopped.