[
https://issues.apache.org/jira/browse/IMPALA-9687?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17261699#comment-17261699
]
ASF subversion and git services commented on IMPALA-9687:
---------------------------------------------------------
Commit 425e424b37f081ba1262f4f39298c6c13910aba8 in impala's branch
refs/heads/master from Akos Kovacs
[ https://gitbox.apache.org/repos/asf?p=impala.git;h=425e424 ]
IMPALA-9687 Improve estimates for number of hosts in Kudu plans
In some cases Kudu plans could contain more hosts than the actual number of
executors.
This commit fixes it by capping the number of hosts at the number of executors,
and determining which executors have local scan ranges.
Testing:
- Ran core tests
Updated Kudu planner tests where the memory estimates changed.
Change-Id: I72e341597e980fb6a7e3792905b942ddf5797d03
Reviewed-on: http://gerrit.cloudera.org:8080/16880
Reviewed-by: Impala Public Jenkins <[email protected]>
Tested-by: Impala Public Jenkins <[email protected]>
> Plans for Kudu can contains hosts > num of Impala nodes
> -------------------------------------------------------
>
> Key: IMPALA-9687
> URL: https://issues.apache.org/jira/browse/IMPALA-9687
> Project: IMPALA
> Issue Type: Bug
> Components: Frontend
> Reporter: Csaba Ringhofer
> Assignee: Akos Kovacs
> Priority: Major
> Labels: kudu, ramp-up
>
> Started a cluster with 1 coordinator and 1 executor:
> {code}
> bin/start-impala-cluster.py -c 1 --use_exclusive_coordinator -s 2
> {code}
> Queries on Kudu tables still contained hosts=3, while all other table types
> had hosts=1.
> {code}
> explain select * from functional_kudu.alltypestiny;
> Result:
> ...
> F00:PLAN FRAGMENT [RANDOM] hosts=3 instances=3
> ...
> {code}
> The numbers in the plan are only estimates, the summary after the query
> contained 1 host.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]