Does iceberg.catalog.warehouse support HDSF NameNode HA

[zhangrainpeng]

presto version is 0.291

[tmc@node4 catalog]$ cat iceberg_hadoop.properties

connector.name=iceberg

iceberg.catalog.type=hadoop

iceberg.catalog.warehouse=hdfs://datacluster/user/iceberg/warehouse

iceberg.hadoop.config.resources=/opt/soft/hadoop-3.4.0/etc/hadoop/core-site.xml,/opt/soft/hadoop-3.4.0/etc/hadoop/hdfs-site.xml

namenode is HA

fs.defaultFS

hdfs://datacluster

CREATE SCHEMA iceberg_hadoop.test_db

[2025-12-02 15:42:41] completed in 101 ms

CREATE TABLE iceberg_hadoop.test_db.players (

id int,

name varchar,

team varchar

)

[2025-12-02 15:43:49] completed in 354 ms

error：

INSERT INTO iceberg_hadoop.test_db.players VALUES(1, 'zrp', 'cmdi')

[2025-12-02 15:44:47] java.net.UnknownHostException: datacluster

[hantangwangd]

This behavior seems to indicate that your coordinator can access HDFS successfully, but the worker nodes are failing to connect. Please make sure the worker nodes can access HDFS as well. Have you added the following HDFS configuration to the etc/catalog/iceberg_hadoop.properties on your worker nodes as well?

iceberg.hadoop.config.resources=/opt/soft/hadoop-3.4.0/etc/hadoop/core-site.xml,/opt/soft/hadoop-3.4.0/etc/hadoop/hdfs-site.xml

[zhangrainpeng]

Thank you for your reply. The catalog configuration for all 4 nodes is the same.

[hantangwangd]

In Iceberg's Hadoop catalog, a successful table creation indicates that the coordinator can write to HDFS successfully. As confirmation, you can execute SHOW CREATE TABLE iceberg_hadoop.test_db.players to verify the table's location and check if the metadata/v1.metadata.json file has been created within it.

Regarding the write failure: since write operations are executed on worker nodes, please double check the configuration on those workers. Ensure that core-site.xml and hdfs-site.xml are properly configured and loaded so that HDFS access is successful.

[zhangrainpeng]

v1.metadata.json