Back Up and Restore Kudu Tables

Kudu tables can be backed up and restored using Spark 3. This process requires a Kudu backup JAR file distributed with Kudu.

Prerequisites

Before performing a backup, ensure the following:

1. ODP version

2. Backup location – HDFS path where backups will be stored.

3. Permissions – The Kudu user must have permission to submit Spark jobs.

4. Tables to back up – You can specify one or multiple tables at the same time.

5. Kudu backup JAR – Located at:

6. List of Kudu masters – Required to connect to the cluster.

Back Up Tables

Run the following Spark command to back up one or more tables:

Parameters:

• --kuduMasterAddresses – Comma-separated list of Kudu master addresses.

• --rootPath – HDFS path where backups will be stored.

• --failOnFirstError – Stops the backup if any table fails.

Restore Tables

Restoring from a backup requires similar inputs as backing up, but with a few key differences.

The --class org.apache.kudu.backup.KuduRestore option specifies that the operation restores tables instead of backing them up.

Parameters:

• --class org.apache.kudu.backup.KuduRestore → Runs the restore job.

• --kuduMasterAddresses → Comma-separated list of Kudu master addresses.

• --rootPath → HDFS path where the backups are stored.

• --failOnFirstError → Ensures the restore stops on the first encountered error.

• foo bar → Placeholder arguments; replace with your specific restore targets.

For more details, see Apache Documentation.