Zoltán Borók-Nagy created HIVE-25226:
----------------------------------------
Summary: Hive changes 'storage_handler' for existing Iceberg table
when hive.engine.enabled is false
Key: HIVE-25226
URL: https://issues.apache.org/jira/browse/HIVE-25226
Project: Hive
Issue Type: Bug
Reporter: Zoltán Borók-Nagy
If Hive writes to an existing Iceberg table but property 'hive.engine.enabled'
is not set, then Hive rewrites the table metadata with different
SerDe/Input/Output format than it had before.
E.g. there's an existing table with the following metadata:
{noformat}
storage_handler |
org.apache.iceberg.mr.hive.HiveIcebergStorageHandler
| SerDe Library: | org.apache.iceberg.mr.hive.HiveIcebergSerDe | NULL |
| InputFormat: | org.apache.iceberg.mr.hive.HiveIcebergInputFormat | NULL |
| OutputFormat: | org.apache.iceberg.mr.hive.HiveIcebergOutputFormat | NULL |
{noformat}
Now when Hive inserts to this table it clears 'storage_handler' and rewrites
the rest:
{noformat}
| SerDe Library: |
org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe | NULL
|
| InputFormat: | org.apache.hadoop.mapred.FileInputFormat
| NULL |
| OutputFormat: | org.apache.hadoop.mapred.FileOutputFormat
| NULL |
{noformat}
This means the table becomes unreadable:
{noformat}
Error: java.io.IOException: java.io.IOException: Cannot create an instance of
InputFormat class org.apache.hadoop.mapred.FileInputFormat as specified in
mapredWork! (state=,code=0)
{noformat}
I think Hive should always set 'hive.engine.enabled' for Iceberg.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
