[
https://issues.apache.org/jira/browse/DRILL-544?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13974208#comment-13974208
]
Kunal Khatua commented on DRILL-544:
------------------------------------
I attempted to run TPC-H5 against a scaleFactor=0.1. (100MB data).
The first run completed with 5 rows returned.
The second run failed with an ArrayIndexOutOfBoundsException. The fact that the
second run failed with this exception and there was an increase in memory size
of the DrillBit process leads me to believe that there is a possible memory
leak.
The following message was obtained from Lilith:
New fragment status was provided to Foreman of memory_use: 1036064
batches_completed: 0
records_completed: 0
state: FAILED
data_processed: 0
handle {
query_id {
part1: 5034909394138843503
part2: -6803985636041355296
}
major_fragment_id: 8
minor_fragment_id: 0
}
error {
error_id: "beb5a920-07c1-4a70-8070-176a9ab1cbaf"
endpoint {
address: "kk127"
user_port: 31010
control_port: 31011
data_port: 31012
}
error_type: 0
message: "Failure while running fragment. < ArrayIndexOutOfBoundsException:[
11570 ]"
}
running_time: 2275863097
> DrillBit is un-responsive after a query with multiple joins is executed
> repeatedly
> ----------------------------------------------------------------------------------
>
> Key: DRILL-544
> URL: https://issues.apache.org/jira/browse/DRILL-544
> Project: Apache Drill
> Issue Type: Bug
> Environment: Centos 6.4
> 96GB RAM
> Reporter: Kunal Khatua
>
> When running a query against a standalone DrillBit with default settings,
> every run of a query with joins results in the memory for the DrillBit
> process increasing. This occurs, irrespective of whether the query runs to
> completion or not.
> After a number of runs, the DrillBit becomes unresponsive, with no CPU
> activity visible with the top command.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
