[
https://issues.apache.org/jira/browse/IMPALA-10036?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17169711#comment-17169711
]
ASF subversion and git services commented on IMPALA-10036:
----------------------------------------------------------
Commit 521c81466f980b95972026a7ac112cfa47a3f854 in impala's branch
refs/heads/master from Tim Armstrong
[ https://gitbox.apache.org/repos/asf?p=impala.git;h=521c814 ]
IMPALA-10036: schedule unpartitioned fragments on random executor
Some plans have unpartitioned fragments that are not the coordinator
fragment. It would be best to avoid scheduling these on a coordinator,
both to reduce coordinator load and to allow using the coordinator
mem estimates for these queries.
Testing:
Extended the dedicated coordinator admission control test with
a query with an unpartitioned fragment. The test failed on master
with dedicated coordinator estimates.
Manually ran the repro with TPC-DS Q23, which is a more complex
query with unpartitioned fragments, and confirmed that could
run with low coordinator mem limits.
Ran exhaustive tests.
Change-Id: I1eee78f24b048968f2fdf91f0da95b78552f8d33
Reviewed-on: http://gerrit.cloudera.org:8080/16272
Reviewed-by: Tim Armstrong <[email protected]>
Tested-by: Tim Armstrong <[email protected]>
> Admission control incorrectly rejecting query based on coordinator limit
> ------------------------------------------------------------------------
>
> Key: IMPALA-10036
> URL: https://issues.apache.org/jira/browse/IMPALA-10036
> Project: IMPALA
> Issue Type: Bug
> Components: Backend
> Affects Versions: Impala 4.0
> Reporter: David Rorke
> Assignee: Tim Armstrong
> Priority: Major
> Attachments: 23_profile.txt, backends-repro.png, impalad.INFO.gz,
> start-ded-coord-cluster.sh, tmp-fair-scheduler.xml, tmp-llama-site.xml
>
>
> Admission control rejected a query with the following error:
> {noformat}
> ERROR: Rejected query from pool root.default: request memory needed 50.00 GB
> is greater than memory available for admission 45.36 GB of
> coordinator-0:22050.
> {noformat}
> The query profile shows the following resource estimates:
> {noformat}
> Per-Host Resource Estimates: Memory=303.36GB
> Dedicated Coordinator Resource Estimate: Memory=136MB
> {noformat}
> The max query memory limit for the pool was configured as 50 GB.
> Based on the coordinator resource estimate of 136 MB the query should have
> easily fit in the available coordinator memory. [~tarmstrong] speculated that
> we might simply be comparing the wrong resource estimate to the coordinator
> limit.
> Profile attached.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
