Upgrading Splice Machine Version 2.0 to Version 2.5

This topic walks you through upgrading your Splice Machine Database from version 2.0 to version 2.5, in these steps:

Step 1: Run a Full Backup

Back up your database, so that you can recover if anything goes amiss.

We also recommend that you create a JSON backup dump of your current 2.0 configuration, which you’ll find here:

http://<your cloudera manager host>:7180/api/v12/cm/deployment

Step 2: If Necessary, Upgrade to a Later Version of 2.0

This upgrade requires you to start with a Version 2.0 release that includes the needed upgrade code, which is any version numbered 2.0.1.1725 or greater. If you’re running an earlier version of Splice Machine v2.0, please follow these steps before starting the upgrade to 2.5:

  1. Deploy version 2.0.1.1725 or later.
  2. Activate and restart hbase.
  3. Verify that the upgrade was successful: You should see this message in the HBase master log: Conglomerates upgrade complete

  4. Wait for hbase to complete startup and make sure you can connect and run some queries.

Recovering from Upgrade Issues

If your upgrade to a more recent 2.0 version fails, you can recover by following these steps:

  1. Deactivate the newly installed Splice Machine 2.0 parcel.
  2. Deploy and activate your previous Splice Machine 2.0 parcel.
  3. Restart HBase.
  4. Flatten (clean) your database.
  5. Restore your database from the backup you created at the beginning of this process.
  6. Restart all cluster services”.

Step 3: Start the Upgrade

The first upgrade step is to prepare your cluster for the new version, following these steps:

  1. Shut down your cluster.
  2. Upgrade the version of JDK running on all cluster nodes to 1.8:

    • Install to /usr/java
    • Delete any other versions of JDK from /usr/java
  3. Delete the Spark service in CM.
  4. Delete all Spark jars from /opt/cloudera/parcels/CDH/jars/

Step 4: Now install and activate Splice Machine, version 2.5.

Activate Splice Machine, but DO NOT restart your cluster yet!

Run the following script on each node in your cluster:

sudo /opt/cloudera/parcels/SPLICEMACHINE/scripts/install-splice-symlinks.sh

Step 5: Make Configuration Changes for Version 2.5

You need to make several updates in your Hadoop configuration so that 2.5 works properly, including:

Please refer to our Cloudera Installation topic (onprem_install_cloudera.html) when making these modifications.

Yarn Updates

You need to add the same two property values to each of four YARN advanced configuration settings.

Add these properties:

Property Name Property Value
yarn.nodemanager.aux-services.spark_shuffle.class org.apache.spark.network.yarn.YarnShuffleService
yarn.nodemanager.aux-services mapreduce_shuffle,spark_shuffle

to each of these YARN settings:

  • Yarn Service Advanced Configuration Snippet (Safety Valve) for yarn-site.xml
  • Yarn Client Advanced Configuration Snippet (Safety Valve) for yarn-site.xml
  • NodeManager Advanced Configuration Snippet (Safety Valve) for yarn-site.xml
  • ResourceManager Advanced Configuration Snippet (Safety Valve) for yarn-site.xml

HBase Updates

Make these changes to the HBase Configuration:

Make these temporary changes to your Hbase Configuration:

  • Set Maximum HBase Client Retries to 400.
  • Set RPC Timeout to 40.

Undo these temporary changes after the upgrade has completed.

Recovering from Upgrade Issues

If your upgrade from version 2.0 to version 2.5 fails, you can recover by following these steps:

  1. Stop all services in CM.
  2. Deactivate the Splice Machine CDH parcel.
  3. Remove the Splice Machine CDH parcel.
  4. Delete the Splice Machine CDH parcel.
  5. Download the Splice Machine 2.0 CDH parcel
  6. Distribute the Splice Machine 2.0 CDH parcel to all nodes.
  7. Activate the Splice Machine 2.0 CDH parcel.