Thanks! Before you replied I tried the following which also helped:   yarn.scheduler.capacity.schedule-asynchronously.maximum-threads=2 yarn.scheduler.capacity.schedule-asynchronously.scheduling-interval-ms=100     ... View more

Hi @Bender    Correct, our cluster is not kerberized  yet.   Is it possible connect SMM without kerberos?   thanks Eduardo ... View more

Hi @apedroso ,   thank you for starting this thread.   In this reply, I will focus on how the YARN RM stores data about historical applications, which can be accessed via the RM Web UI.   The RM keeps data about the applications in its state store [1]. It can be LeveldbRMStateStore, FileSystemRMStateStore or ZKRMStateStore.   We recommend using ZKRMStateStore (this is what we use in YARN HA as well), because it is a more robust implementation. For example, you can migrate in RM HA standby RM while the active RM is still running and keep the state-store intact.   Because the RM Web UI is reading the data from the state-store, it is independent of the presence or lack of YARN Application Logs.   What are your exact migration steps? Do I understand correctly that you upgrade your cluster to CDP or do you need to move services to a new cluster, please?   [1] Please read section for "yarn.resourcemanager.store.class" in https://hadoop.apache.org/docs/r3.1.2/hadoop-yarn/hadoop-yarn-site/ResourceManagerRestart.html ... View more