Flink on yarn of HA

The first step in the configuration file flink-conf.yaml

high-availability: ZOOKEEPER
high-availability.cluster-id: eve_flink_1
high-availability.storageDir: hdfs:///flink/checkpoints
high-availability.zookeeper.path.root: /flink
high-availability.zookeeper.quorum: test01:2181,test02:2181,test03:2181

yarn.application-attempts: 2

The second step in the configuration zoo.cfg Flink conf directory and add some configuration, delete the original

server.1=test01:2888:3888
server.2=test01:2888:3888
server.3=test01:2888:3888

The third step is to start Flink Job, can make the session mode, it can be flink run mode.

The fourth step verification HA

1. Firstly yarn of web ui find flink job submission, and then find the task jobmanager on which machine, then the machine up to kill off jobmanager. If session mode process name for YarnSessionClusterEntrypoint , if it is flink run way process name for YarnJobClusterEntrypoint .
2. The second step in the Job appattempt see whether updated. Results are as follows