[ 
https://issues.apache.org/jira/browse/HIVE-26409?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Sourabh Badhya updated HIVE-26409:
----------------------------------
    Description: NO_TXN operation type has to be assigned while acquiring locks 
for the table in global locks because this will not add records to 
COMPLETED_TXN_COMPONENTS table. Currently the record introduced by this lock 
had writeId as NULL. There was a situation which lead to large number of 
records in COMPLETED_TXN_COMPONENTS table because these records had writeId as 
NULL which weren't deleted up by AcidHouseKeeperService/Cleaner and this 
subsequently lead to OOM errors.  (was: A unique writeId has to be assigned for 
the table in global locks because this will help in cleanup of records in 
COMPLETED_TXN_COMPONENTS table. Currently the writeId is NULL for this table. 
There was a situation which lead to large number of records in 
COMPLETED_TXN_COMPONENTS table because these records had writeId as NULL and 
this subsequently lead to OOM errors.)

> Assign NO_TXN operation type for table in global locks for scheduled queries
> ----------------------------------------------------------------------------
>
>                 Key: HIVE-26409
>                 URL: https://issues.apache.org/jira/browse/HIVE-26409
>             Project: Hive
>          Issue Type: Bug
>            Reporter: Sourabh Badhya
>            Assignee: Sourabh Badhya
>            Priority: Major
>              Labels: pull-request-available
>          Time Spent: 20m
>  Remaining Estimate: 0h
>
> NO_TXN operation type has to be assigned while acquiring locks for the table 
> in global locks because this will not add records to COMPLETED_TXN_COMPONENTS 
> table. Currently the record introduced by this lock had writeId as NULL. 
> There was a situation which lead to large number of records in 
> COMPLETED_TXN_COMPONENTS table because these records had writeId as NULL 
> which weren't deleted up by AcidHouseKeeperService/Cleaner and this 
> subsequently lead to OOM errors.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to