When Apache Iceberg creates tables, it stores the current NameNode address in the
metadata. If HDFS is running without High Availability (HA), Iceberg stores the
NameNode's hostname and port. If HDFS is configured with HA, it stores the
nameservice ID (example, nameservice1).
Consequently, modifying the HDFS configuration such as transitioning from a non-HA to
an HA setup or reverting from HA to non-HA, changes the NameNode address and
invalidates the stored address, rendering the existing data unreadable. In these
scenarios, you must export and re-import any existing Iceberg data after the HDFS
changes are finalized to maintain data accessibility.
For more information, see Create table feature using
Apache Iceberg.
