MR3 in standalone mode allows Hive on MR3 to run without requiring a resource manager such as Hadoop or Kubernetes. The user can install Hive on MR3 virtually in any type of cluster regardless of the availability of Hadoop or Kubernetes and the version of Java installed in the system. A storage cluster hosting HDFS or S3 (or S3-compatible storage) is, however, still required.
The main advantage of standalone mode is the simplicity of installing and operating Hive on MR3. To install Hive on MR3 in standalone mode, it suffices to download an MR3 distribution built for standalone mode, configure Hive on MR3 by updating a few configuration files, and copy the updated MR3 distribution to every node. Hence installing Hive on MR3 is similar to installing Trino/Presto and Spark in standalone mode.
Since Hive on MR3 in standalone mode does not use a resource manager, it does not directly support a few key features such as high availability, DAGAppMaster recovery, and autoscaling. This is a price to pay for using standalone mode, in which the user is responsible for launching and terminating every component.
For asking any questions, please visit MR3 Google Group or join MR3 Slack.
