[jira] [Created] (HIVE-24796) [HS2] Enhance DriverTxnHandler.isValidTxnListState logic to include tableId comparison

2021-02-18 Thread Kishen Das (Jira)
Kishen Das created HIVE-24796:
-

 Summary: [HS2] Enhance DriverTxnHandler.isValidTxnListState logic 
to include tableId comparison
 Key: HIVE-24796
 URL: https://issues.apache.org/jira/browse/HIVE-24796
 Project: Hive
  Issue Type: Sub-task
  Components: HiveServer2
Reporter: Kishen Das


In HS2, after query compilation phase, we acquire lock in DriverTxnHandler.

isValidTxnListState and later ensure there are no conflicting transactions by 
using driverContext.getTxnManager().getLatestTxnIdInConflict(); . This doesn't 
work well, if there are external entities that drop and recreate the table with 
the same name. So, we should also make sure the tableId itself is not changed, 
after lock has been acquired. This Jira is to enhance the 
DriverTxnHandler.isValidTxnListState logic to include tableId comparison. 



--
This message was sent by Atlassian Jira
(v8.3.4#803005)


[jira] [Created] (HIVE-24795) Cache tableId in SessionState

2021-02-18 Thread Kishen Das (Jira)
Kishen Das created HIVE-24795:
-

 Summary: Cache tableId in SessionState 
 Key: HIVE-24795
 URL: https://issues.apache.org/jira/browse/HIVE-24795
 Project: Hive
  Issue Type: Sub-task
  Components: HiveServer2
Reporter: Kishen Das


Please go through https://issues.apache.org/jira/browse/HIVE-24794 to 
understand why this is required. Its basically to handle a corner case, in 
which a table gets dropped and recreated with the same name, after we gather 
information about all the tables and we acquire the lock. 



--
This message was sent by Atlassian Jira
(v8.3.4#803005)


[jira] [Created] (HIVE-24794) Populate tableId in the response of get_valid_write_ids API

2021-02-18 Thread Kishen Das (Jira)
Kishen Das created HIVE-24794:
-

 Summary: Populate tableId in the response of get_valid_write_ids 
API
 Key: HIVE-24794
 URL: https://issues.apache.org/jira/browse/HIVE-24794
 Project: Hive
  Issue Type: Sub-task
  Components: HiveServer2
Reporter: Kishen Das


In HS2, after query compilation phase, we acquire lock in DriverTxnHandler.

isValidTxnListState and later ensure there are no conflicting transactions by 
using driverContext.getTxnManager().getLatestTxnIdInConflict(); . This doesn't 
work well, if there are external entities that drop and recreate the table with 
the same name. So, we should also make sure the tableId itself is not changed, 
after lock has been acquired. This Jira is to enhance getValidWriteIdList to 
return tableId as well. Idea is to cache the tableId in SessionState and later 
compare it with what getValidWriteIdList returns. If the table was dropped and 
recreated, the tableId will not match and we have to recompile the query. 
Caching the tableId in SessionState will be done as part of a separate Jira. 



--
This message was sent by Atlassian Jira
(v8.3.4#803005)


[jira] [Created] (HIVE-24793) Compiler probe MJ selection algorithm fallback

2021-02-18 Thread Panagiotis Garefalakis (Jira)
Panagiotis Garefalakis created HIVE-24793:
-

 Summary: Compiler probe MJ selection algorithm fallback
 Key: HIVE-24793
 URL: https://issues.apache.org/jira/browse/HIVE-24793
 Project: Hive
  Issue Type: Sub-task
Reporter: Panagiotis Garefalakis
Assignee: Panagiotis Garefalakis


As per discussion on #1286 current probe MJ selection algorithm:

* selects best MJ candidate (currently based on the distinct row ratio)
* does some further processing - which may bail out

Bailing out for the best candidate doesn't necessarily mean that we can not use 
a less charming candidate. The extra compilation can be wrapped into to for 
loop instead of selecting the best candidate the first part could be 
implemented as a priority logic



--
This message was sent by Atlassian Jira
(v8.3.4#803005)


[jira] [Created] (HIVE-24792) Potential thread leak in Operation

2021-02-18 Thread Zhihua Deng (Jira)
Zhihua Deng created HIVE-24792:
--

 Summary: Potential thread leak in Operation
 Key: HIVE-24792
 URL: https://issues.apache.org/jira/browse/HIVE-24792
 Project: Hive
  Issue Type: Bug
  Components: HiveServer2
Reporter: Zhihua Deng


The _scheduledExecutorService_  in _Operation_ does not shut down after 
scheduling delay operationlog cleanup, which may result to thread leak in 
hiveserver2...



--
This message was sent by Atlassian Jira
(v8.3.4#803005)