[
https://issues.apache.org/jira/browse/DRILL-4274?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15102123#comment-15102123
]
Deneche A. Hakim commented on DRILL-4274:
-----------------------------------------
Many functional tests that use a "wide table" fail with the following error
when run on 32 core machines (or using planner.width.max_per_node=23):
{noformat}
/root/drillAutomation/framework-master/framework/resources/Functional/data-shapes/wide-columns/5000/1000rows/parquet/q163_DRILL-2046.q
Query:
select ws1.* from widestrings ws1 INNER JOIN widestrings ws2 on
ws1.str_var=ws2.str_var_null_empty
Failed with exception
java.sql.SQLException: RESOURCE ERROR: One or more nodes ran out of memory
while executing the query.
Unable to allocate sv2 for 1000 records, and not enough batchGroups to spill.
batchGroups.size 0
spilledBatchGroups.size 0
allocated memory 52520576
allocator limit 46684427
Fragment 0:0
{noformat}
In this case the sort exceeds it's limit on the first data batch, as you can
see from the message context, the batch wasn't added to batchGroups yet.
> ExternalSort doesn't always handle low memory condition well, failing
> execution instead of spilling in some cases
> -----------------------------------------------------------------------------------------------------------------
>
> Key: DRILL-4274
> URL: https://issues.apache.org/jira/browse/DRILL-4274
> Project: Apache Drill
> Issue Type: Bug
> Components: Execution - Relational Operators
> Reporter: Jacques Nadeau
> Assignee: Jacques Nadeau
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
