[
https://issues.apache.org/jira/browse/HIVE-9839?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14344992#comment-14344992
]
Nemon Lou commented on HIVE-9839:
---------------------------------
The piece of code in org.apache.hive.service.cli.session.HiveSessionImpl.java
that causing this issue:
{code}
private OperationHandle executeStatementInternal(String statement,
Map<String, String> confOverlay,
370 boolean runAsync)
371 throws HiveSQLException {
372 acquire(true);
373
374 OperationManager operationManager = getOperationManager();
375 ExecuteStatementOperation operation = operationManager
376 .newExecuteStatementOperation(getSession(), statement,
confOverlay, runAsync);
377 OperationHandle opHandle = operation.getHandle();
378 try {
379 operation.run();
380 opHandleSet.add(opHandle);
381 return opHandle;
382 } catch (HiveSQLException e) {
383 // Cleanup opHandle in case the query is synchronous
384 // Async query needs to retain and pass back the opHandle for
error reporting
385 if (!runAsync) {
386 operationManager.closeOperation(opHandle);
387 }
388 throw e;
389 } finally {
390 release(true);
391 }
392 }
{code}
When a HiveSQLException throw from line 379,the opHandle is not closed due to
async mode and is not been put into opHandleSet for future clean up.
Async query is the default query mode in hiveserver2 clients.
> HiveServer2 leaks OperationHandle on failed async queries
> ---------------------------------------------------------
>
> Key: HIVE-9839
> URL: https://issues.apache.org/jira/browse/HIVE-9839
> Project: Hive
> Issue Type: Bug
> Components: HiveServer2
> Affects Versions: 0.14.0, 0.13.1, 1.0.0
> Reporter: Nemon Lou
>
> Using beeline to connect to HiveServer2.And type the following:
> drop table if exists table_not_exists;
> select * from table_not_exists;
> There will be an OperationHandle object staying in HiveServer2's memory for
> ever even after quit from beeline .
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
