Hi, Jake In Flink 1.14, we recommend using "-t yarn-per-job"[1] for starting per-job cluster. Regarding your issue, I could not reproduce it with the Wordcount example. However, I think this is not the right way for Flink's SQL client, which might be the root cause of your issue. Would you like to take a look at [2]?
[1] https://ci.apache.org/projects/flink/flink-docs-master/docs/deployment/resource-providers/yarn/#per-job-cluster-mode [2] https://ci.apache.org/projects/flink/flink-docs-master/docs/dev/table/sqlclient/ Best, Yangze Guo On Tue, Oct 12, 2021 at 10:52 AM Caizhi Weng <tsreape...@gmail.com> wrote: > > Hi! > > yarn-cluster is the mode for a yarn session cluster, which means the cluster > will remain even after the job is finished. If you want to finish the Flink > job as well as the yarn job, use yarn-per-job mode instead. > > Jake <ft20...@qq.com> 于2021年10月9日周六 下午5:53写道: >> >> Hi >> >> When submit job in yarn-cluster model, flink job finish but yarn job not >> exit. What should I do? >> >> Submit command: >> >> /opt/app/flink/flink-1.14.0/bin/flink run -m yarn-cluster >> ./flink-sql-client.jar --file dev.sql