Manual Rolling Upgrade Procedure

This procedure assumes that the cluster being upgraded is running release 6.1.0:

1. Install the latest core 6.1.0 patch on all nodes of the release 6.1.0 cluster. The latest core 6.1.0 patch must be installed on all nodes before you start the rolling upgrade.

2. Download the archive file from https://package.mapr.com/releases/.

3. Extract the archive file. When you upgrade each package, be sure to specify the full path to the files in this local directory:

   tar -xzvf <archive file>

4. Stop CLDB if it is running on the node, before putting that node in maintenance mode. Else, the maintenance mode operation is not permitted:

   maprcli node services -name cldb -action stop -nodes <hostname>

5. Set the node to maintenance mode:

   sudo maprcli node maintenance -nodes <hostname> -timeoutminutes 30

6. Notify the CLDB that the node is going to be upgraded:

   sudo maprcli notifyupgrade start -node <hostname>

7. Stop Warden:

   sudo service mapr-warden stop

8. If ZooKeeper is installed on the node, stop ZooKeeper:

   service mapr-zookeeper stop

9. Check to ensure that all services are stopped:

   jps 
   46343 Jps
   3607 -- process information unavailable
   (Nothing running here)
   

10. Use the following command to view the Java alternatives menu, and set Java to JDK 11:

    sudo update-alternatives --config java

11. Remove any patches installed on the node.

    • RHEL/CentOS or SUSE: sudo rpm -e mapr-patch
    • Ubuntu: sudo apt-get -y remove mapr-patch

12. Upgrade each data-fabric package on the node based on the defined package upgrade order by running this command for each package:

    • RHEL/CentOS or SUSE: yum upgrade <packagename>
    • Ubuntu: apt-get install --only-upgrade -o Dpkg::Options::="--force-overwrite" mapr-fileserver mapr-core mapr-cldb
      Note: During the upgrade process on Ubuntu, the system displays dpkg warnings about overwriting. You can ignore these warnings.
13. Verify that the following packages, which are obsolete in release 6.2, are removed:
    • Remove mapr-mapreduce2 if it is present
    • Remove mapr-hadoop-core if you are upgrading a fileserver node that does not have the ResourceManager, NodeManager, or History Server. On such a node, only mapr-hadoop-util is needed after upgrading.
14. Run configure.sh with the -disableSsl option:

    /opt/mapr/server/configure.sh -R -disableSsl

    Disabling SSL prevents the upgraded node from attempting to use SSL with nodes that have not been upgraded. Release 6.2.0 nodes use SSL, but release 6.1.0 nodes do not. After all the ZooKeeper nodes have been upgraded, the SSL feature can be turned on by stopping the cluster, running configure.sh -R on all nodes and restarting the cluster.
15. Ensure that every node has the following files in the /opt/mapr/conf directory:
    • maprserverticket
    • ssl_keystore
    • ssl_keystore.p12
    • ssl_keystore.pem
    • ssl_truststore
    • ssl_truststore.p12
    • ssl_truststore.pem
16. If ZooKeeper is installed on the node, start ZooKeeper:

    service mapr-zookeeper start

17. Start Warden:

    sudo service mapr-warden start

18. Check that the CLDB is running. If output is displayed, the CLDB is running. If not, start CLDB:

    maprcli node list

19. Unset maintenance mode on the node, notify the CLDB about the upgraded version, and about the finished status of the upgrade process.

    sudo maprcli node maintenance -nodes <hostname> -timeoutminutes 0
    sudo maprcli config save -values {mapr.targetversion:"'cat /opt/mapr/MapRBuildVersion'"}
    sudo maprcli notifyupgrade finish -node <hostname>

20. Wait for the containers to synchronize. Then run the following command, and check that there is no output.

    /opt/mapr/server/mrconfig info containers resync local

    No output signifies that the containers are synchronized.