At the end of upgrading all worker nodes, then run the same set of validation tests to ensure that the cluster is functional.
At this point, validate the upgrade process by running the following tests. Make sure that these jobs execute successfully. If they succeed, proceed with rolling upgrade.
MapReduce job
Hive/Tez job
hadoop dfsadmin -report
yarn –report
hbase status
See Appendix A for more information.
If the upgrade process fails, follow the steps in Downgrading the Cluster, starting with "Downgrade core slave nodes."