[
https://issues.apache.org/jira/browse/DRILL-7037?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16767174#comment-16767174
]
Vitalii Diravka commented on DRILL-7037:
----------------------------------------
[[email protected]] Thanks for the stack trace. It clearly shows that
issue come from the Calcite. Drill can't build the plan for this huge query.
Calcite HEP planner creates a big DAG with a RelNode vertices. It uses
DEPTH_FIRST algorithm to apply different rules for optimization and then
recursively process children to obtain the best logical plan. And HEAP memory
is used for that whole process.
Calcite is a SQL planner for many engines. Therefore looks like it should be
solved in Calcite project.
You can ask Calcite [community|https://calcite.apache.org/community/#help]
about it and then [create ticket
|https://issues.apache.org/jira/projects/CALCITE].
> Apache Drill Crashes when a 50mb json string is queried via the REST API
> provided
> ---------------------------------------------------------------------------------
>
> Key: DRILL-7037
> URL: https://issues.apache.org/jira/browse/DRILL-7037
> Project: Apache Drill
> Issue Type: Bug
> Components: Client - HTTP
> Affects Versions: 1.14.0
> Environment: Windows 10
> 24GB RAM
> 8 Cores
> Used the REST API call to query drill
> Reporter: Ayush Sharma
> Priority: Blocker
>
> Apache Drill crashes with OutofMemoryException (24GB RAM) when a REST API
> call is made by supplying a json of size 50MB in the query paramater of the
> REST API.
> The REST API even crashes for a 10MB query (16GB RAM) and works with a 5MB
> query.
> This is a blocker for us and will need immediate remediation.
> We are also not aware of any sys.options which might bring the HEAP size down
> drastically or currently making it go up.
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
