[
https://issues.apache.org/jira/browse/HIVE-26121?focusedWorklogId=753312&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-753312
]
ASF GitHub Bot logged work on HIVE-26121:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 06/Apr/22 11:27
Start Date: 06/Apr/22 11:27
Worklog Time Spent: 10m
Work Description: deniskuzZ commented on code in PR #3181:
URL: https://github.com/apache/hive/pull/3181#discussion_r843829613
##########
ql/src/java/org/apache/hadoop/hive/ql/lockmgr/DbTxnManager.java:
##########
@@ -574,30 +571,24 @@ public void rollbackTxn() throws LockException {
if (!isTxnOpen()) {
throw new RuntimeException("Attempt to rollback before opening a
transaction");
}
- stopHeartbeat();
-
try {
- lockMgr.clearLocalLockRecords();
+ clearLocksAndHB();
LOG.debug("Rolling back " + JavaUtils.txnIdToString(txnId));
-
- // Re-checking as txn could have been closed, in the meantime, by a
competing thread.
- if (isTxnOpen()) {
- if (replPolicy != null) {
- getMS().replRollbackTxn(txnId, replPolicy, TxnType.DEFAULT);
- } else {
- getMS().rollbackTxn(txnId);
- }
+
+ if (replPolicy != null) {
Review Comment:
marked as @NotThreadSafe
Issue Time Tracking
-------------------
Worklog Id: (was: 753312)
Time Spent: 20m (was: 10m)
> Hive transaction rollback should be thread-safe
> -----------------------------------------------
>
> Key: HIVE-26121
> URL: https://issues.apache.org/jira/browse/HIVE-26121
> Project: Hive
> Issue Type: Task
> Reporter: Denys Kuzmenko
> Priority: Major
> Labels: pull-request-available
> Time Spent: 20m
> Remaining Estimate: 0h
>
> When Hive query is being interrupted via cancel request, both the background
> pool thread (HiveServer2-Background) executing the query and the HttpHandler
> thread (HiveServer2-Handler) running the HiveSession.cancelOperation logic
> will eventually trigger the below method:
> {code}
> DriverTxnHandler.endTransactionAndCleanup(boolean commit)
> {code}
> Since this method could be invoked concurrently we need to synchronize access
> to it, so that only 1 thread would attempt to abort the transaction and stop
> the heartbeat.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
