You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@drill.apache.org by "huntersjm (JIRA)" <ji...@apache.org> on 2015/11/19 02:40:11 UTC

[jira] [Commented] (DRILL-4100) DRILL JOIN STREAMING AGG ERROR

    [ https://issues.apache.org/jira/browse/DRILL-4100?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15012565#comment-15012565 ] 

huntersjm commented on DRILL-4100:
----------------------------------

That means `streaming agg` operation in plans tuncates the output, you can get that from the Operator Profiles:
{noformat}
01-xx-00 - HASH_PARTITION_SENDER
Minor Fragment	Setup Time	Process Time	Wait Time	Max Batches	Max Records Peak Memory
01-00-00	0.000s	0.301s	0.028s	3	65,536	90KB
01-01-00	0.000s	0.406s	0.028s	3	65,536	91KB
01-xx-01 - STREAMING_AGGREGATE
Minor Fragment	Setup Time	Process Time	Wait Time	Max Batches	Max Records Peak Memory
01-00-01	0.174s	0.564s	0.000s	29	1,812,457	1MB
01-01-01	0.497s	0.553s	0.000s	29	1,812,973	1MB
{noformat}
After streaming agg, 1,812,457 records becomes 65,536 records

> DRILL JOIN STREAMING AGG ERROR
> ------------------------------
>
>                 Key: DRILL-4100
>                 URL: https://issues.apache.org/jira/browse/DRILL-4100
>             Project: Apache Drill
>          Issue Type: Bug
>          Components: Functions - Drill
>    Affects Versions: 1.1.0
>            Reporter: huntersjm
>
> SQL LIKE :
> {code:sql}
> select t1.a1 from (
> select a1,count(a2) as total from t  group by a1
> ) as a join (
> select a1,count(a2) as total from t where a3 = 'true' group by a1 order by total desc limit 100
> ) as f on t2.a1 = t1.a1 
> {code}
> GET ERROR RESULT
> PART OF Operator Profiles:
> 01-xx-00 - HASH_PARTITION_SENDER
> Minor Fragment	Setup Time	Process Time	Wait Time	Max Batches	Max Records	Peak Memory
> 01-00-00	0.000s	0.301s	0.028s	3	65,536	90KB
> 01-01-00	0.000s	0.406s	0.028s	3	65,536	91KB
> 01-xx-01 - STREAMING_AGGREGATE
> Minor Fragment	Setup Time	Process Time	Wait Time	Max Batches	Max Records	Peak Memory
> 01-00-01	0.174s	0.564s	0.000s	29	1,812,457	1MB
> 01-01-01	0.497s	0.553s	0.000s	29	1,812,973	1MB



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)