@dsun During the upgrade process, a component is supposed to be restarted after the hdp-select command has been run so it will pick up the new binaries. However, the component needs to shut down and start up after the hdp-select command has been run. That way it will report to Ambari that it's version has changed and what it's current state is. In the event that you get stuck (as you did) during the upgrade you can unwind the versioning with a process like this: Make all pieces of the component are running Run `hdp-select set` command on all nodes in the cluster to set the new version. Make sure you get all of the pieces for the component (e.g. hadoop-hdfs-namenode, hadoop-ndfs-journalnode, etc.) Restart all processes for the component Verify that the O/S processes are running with the proper version of jar files Lather, rinse, and repeat for all components in the cluster Once you have successfully gotten everything restarted with the proper bits, you should be able to manually finalize the upgrade with the following command to the Ambari Server: ambari-server set-current --cluster=<custername> --version-display-name=HDP-2.6.2.0 If you get an error that components are not upgraded, you can check the components and hosts again. If everything seems ok, then you may need to tweak a table in the database. I ran into this when Atlas did not properly report the upgraded version to Ambari. NOTE: THIS SHOULD BE DONE WITH THE GUIDANCE OF HORTONWORKS SUPPORT ONLY ambari=> SELECT h.host_name, hcs.service_name, hcs.component_name, hcs.version FROM hostcomponentstate hcs JOIN hosts h ON hcs.host_id = h.host_id ORDER BY hcs.version, hcs.service_name, hcs.component_name, h.host_name;
host_name | service_name | component_name | version
----------------------------------+----------------+-------------------------+-------------
scregione1.field.hortonworks.com | ATLAS | ATLAS_CLIENT | 2.6.1.0-129
scregionm0.field.hortonworks.com | ATLAS | ATLAS_CLIENT | 2.6.1.0-129
scregionm1.field.hortonworks.com | ATLAS | ATLAS_CLIENT | 2.6.1.0-129
scregionm2.field.hortonworks.com | ATLAS | ATLAS_CLIENT | 2.6.1.0-129
scregionw0.field.hortonworks.com | ATLAS | ATLAS_CLIENT | 2.6.1.0-129
scregionw1.field.hortonworks.com | ATLAS | ATLAS_CLIENT | 2.6.1.0-129
scregionm0.field.hortonworks.com | ATLAS | ATLAS_SERVER | 2.6.1.0-129
scregionm1.field.hortonworks.com | DRUID | DRUID_BROKER | 2.6.2.0-205
scregionm1.field.hortonworks.com | DRUID | DRUID_COORDINATOR | 2.6.2.0-205
scregionw0.field.hortonworks.com | DRUID | DRUID_HISTORICAL | 2.6.2.0-205
scregionw1.field.hortonworks.com | DRUID | DRUID_HISTORICAL | 2.6.2.0-205
scregionw0.field.hortonworks.com | DRUID | DRUID_MIDDLEMANAGER | 2.6.2.0-205
scregionw1.field.hortonworks.com | DRUID | DRUID_MIDDLEMANAGER | 2.6.2.0-205
scregionm2.field.hortonworks.com | DRUID | DRUID_OVERLORD | 2.6.2.0-205
scregionm2.field.hortonworks.com | DRUID | DRUID_ROUTER | 2.6.2.0-205
scregionm2.field.hortonworks.com | DRUID | DRUID_SUPERSET | 2.6.2.0-205
scregione1.field.hortonworks.com | HBASE | HBASE_CLIENT | 2.6.2.0-205
scregionm0.field.hortonworks.com | HBASE | HBASE_CLIENT | 2.6.2.0-205
scregionm1.field.hortonworks.com | HBASE | HBASE_CLIENT | 2.6.2.0-205
. . . After verifying that you have, indeed, upgraded the components, a simple update command will set the proper version for the erroneous components and allow you to finalize the upgrade: ambari=> update hostcomponentstate set version='2.6.2.0-205' where component_name = 'ATLAS_CLIENT';
UPDATE 6
ambari=> update hostcomponentstate set version='2.6.2.0-205' where component_name = 'ATLAS_SERVER';
UPDATE 1
After cycling the Ambari Server, you should be able to finalize: [root@hostname ~]# ambari-server set-current --cluster=<cluster> --version-display-name=HDP-2.6.2.0
Using python /usr/bin/python
Setting current version...
Enter Ambari Admin login: <username>
Enter Ambari Admin password:
Current version successfully updated to HDP-2.6.2.0
Ambari Server 'set-current' completed successfully.
... View more
