@Michael Bronson, What is the value of dfs.ha.namenodes.{ha-cluster-name} in your hdfs-site.xml You can get the {ha-cluster-name} from fs.defaultFS from core-site.xml Assuming fs.defaultFS is hdfs://hortonworks. hortonworks is the ha-cluster-name. Thanks, Aditya ... View more

@Michael Bronson, This is because of dependency between the services. These are defined in role_command_order.json. If there is no dependency for the component it can run in parallel between the nodes. However in a particular node the execution will be sequential. You can check the role_command_order.json for different services and their dependencies. Just run the command in ambari server to find all the json files. find /var/lib/ambari-server/ -iname role_command_order.json Thanks, Aditya ... View more

@Kaliyug Antagonist, can you try installing hadooplzo package in all the nodes and try restarting the services yum install -y hadooplzo hadooplzo-native Thanks, Aditya ... View more

@Ashikin, 1)Can you try moving the file /usr/hdp/2.6.3.0-235 to some other folder. Make sure that only 2.5.3.0-37 folder is there in /usr/hdp folder. 2) Then run hdp-select set all 2.5.3.0-37 3)Now run hdp-select versions. Should return only 2.5.3.0-37 4) Restart spark history If it still fails . Then try the manual steps for creating the tar file which i have mentioned above. For the failure which you have mentioned while running the manual steps, you can try creating these folders manually hdfs dfs -mkdir /hdp/apps/2.5.3.0-37/ hdfs dfs -mkdir /hdp/apps/2.5.3.0-37/spark2 hdfs dfs -chown -R hdfs:hdfs /hdp/apps/2.5.3.0-37 ... View more

@Michael Bronson, This looks like a permission issue. The /spark2-history should belong to the spark user. You can change it as below hdfs dfs -chown spark /spark2-history hdfs dfs -chown spark /spark-history Thanks, Aditya ... View more