> For the complete documentation index, see [llms.txt](https://zjffdu.gitbook.io/flink-on-zeppelin/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://zjffdu.gitbook.io/flink-on-zeppelin/execution-mode/yarn-application-mode.md).

# yarn-application mode

Application mode is a new execution mode which is introduced after Flink 1.11. The purpose of application mode is to mitigate the resource pressure on the client side. yarn-application mode would launch a Flink cluster on YARN, where the main() method of the application jar gets executed on the JobManager in YARN.

## Architecture

The is the original yarn mode where Flink Interpreter would run in the same host with Zeppelin server host.

![](/files/-MeJ4sc3FK7EuPDfItfu)

This is yarn-application mode where Flink Interpreter would run in JobManager.

![](/files/-MeJ5-K90iPVldjC0wU1)

## Configuration

It is very to easy to configure `yarn-application` mode, very similar as `yarn` mode

* Configure `FLINK_HOME`
* Configure `flink.execution.mode` as `yarn-application`
* Make sure hadoop client installed, and put hadoop in `PATH`.  Because flink itself doesn't ship hadoop jars, 为Flink本身不带hadoop相关依赖（你可以把flink-shaded-hadoop 拷贝到flink lib下），when zeppelin run flink, it would calll command `hadoop classpath`，then put all the jars under `CLASSPATH`

## Community

Join Zeppelin community to discuss with others 

{% embed url="<http://zeppelin.apache.org/community.html>" %}