diqiu50 opened a new issue, #14560:
URL: https://github.com/apache/gravitino/issues/14560
### Version
main branch
### Describe what's wrong
Several Glue catalog operations on Hive tables, schemas and partitions lose
data or accept requests Glue cannot serve:
- Altering a Hive table rebuilds it from the Gravitino view, dropping Glue
fields such as owner, retention, partition keys, SerDe parameters and view
text. Concurrent alters overwrite each other.
- Altering a schema drops its `LocationUri` and other database fields;
`location` is stored as a parameter instead of `LocationUri`.
- Hive tables are created as Parquet when no format is given (Glue/Athena
default is text), and partial or mismatched input/output/SerDe classes are
accepted.
- Non-trailing or non-identity partition columns, unsupported distributions,
sort orders and auto-increment columns are accepted, then fail or are silently
changed.
- Column names are treated as case-sensitive, but Glue stores them in lower
case, so duplicate or mismatched names reach Glue.
- Partition values are mapped by position, not name; special characters in
values are not escaped; added partitions don't inherit the table storage
descriptor.
- `double precision`, bare `decimal` and unsigned types are not handled
correctly; invalid struct field names are accepted.
### Error message and/or stacktrace
```
ALTER TABLE t SET PROPERTIES ('k'='v') -- on a table created in Athena
-> partition keys and SerDe parameters of t are gone in Glue
```
### How to reproduce
1. Create a partitioned Hive table in Glue (e.g. via Athena) and load it in
a Gravitino Glue catalog.
2. Alter any property or column through Gravitino.
3. Check the table in Glue: fields not modeled by Gravitino are lost.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]