[HIVE-11934] Transaction lock retry logic results in infinite loop

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 1.2.1
• Fix Version/s: 1.3.0
• Component/s: HiveServer2, Transactions
• Assignee: Eugene Koifman Assign to me
• Reporter: Steve Howard
• Created: 23/Sep/15 16:36
• Updated: 01/Oct/19 22:07
• Resolved: 01/Oct/15 00:45

Description

We reset the deadlock count to 0 every time the lock() method is called in org.apache.hadoop.hive.metastore.txn.TxnHandler, so the ten count is never reached in checkRetryable().

We should let checkRetryable handle the deadlock count.

public LockResponse lock(LockRequest rqst)
throws NoSuchTxnException, TxnAbortedException, MetaException
{
>>> this.deadlockCnt = 0; <<<
try
{
Connection dbConn = null;
try
{

Issue Links

• blocks: HIVE-11916 TxnHandler.getOpenTxnsInfo() and getOpenTxns() may produce inconsistent result
• relates to: HIVE-10500 Repeated deadlocks in underlying RDBMS cause transaction or lock failure