nevzheng opened a new issue, #14049: URL: https://github.com/apache/gravitino/issues/14049
### Describe the subtask Tests run Hive SQL, HDFS commands and readiness checks through `docker exec` CLIs. Each `hive -e` call starts a JVM with its own embedded metastore, which costs seconds per statement and per readiness poll. Replace them with in-process clients: - Hive SQL over JDBC to the HiveServer2 that already runs in the container. - Hive readiness through the NameNode's JMX and WebHDFS plus the metastore and HiveServer2 ports. - Kerberos HDFS commands through Hadoop `FileSystem`, database grants over JDBC, LocalStack setup over its HTTP API. The Hive image runs everything as `HADOOP_USER_NAME` (`anonymous` in tests), which may not impersonate users, so HiveServer2 sessions fail. Until the image allows it, the test container adds the proxy-user rule before Hadoop starts. Locally (3 runs, medians): catalog-hive 380 s → 235 s, `CatalogHiveViewIT` 122 s → 37 s, Hive readiness 8 s → 0.5 s per start. ### Parent issue #13900 ### Order and PRs Blocked by #14003. PRs: two PRs stacked on #14013, linked here when opened. The image changes (proxy-user rule and warehouse directory in `start.sh`) follow in a separate issue. -- 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]
