[
https://issues.apache.org/jira/browse/HIVE-7992?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mostafa Mokhtar updated HIVE-7992:
----------------------------------
Description:
When StatsRulesProcFactory overflows it sets data size to 0 and as a result the
Vertex will ask for a single task, this results in a fairly slow running query,
most likely the overflow is a result of higher than usual number of rows.
The class should detect an overflow and set a flag when an overflow occurs, if
an overflow occurs StatsRulesProcFactory should request the maximum number of
tasks for the vertex.
was:
When StatsRulesProcFactory overflows it sets data size to 0 and as a result the
Vertex will ask for a single task, this results in a fairly slow running query,
most likely the overflow is a result of higher than usual number of rows.
The class should detect an overflow and set a flag when an overflow occurs, if
an overflow occurs request the maximum number of tasks for the vertex.
> StatsRulesProcFactory should gracefully handle overflows
> --------------------------------------------------------
>
> Key: HIVE-7992
> URL: https://issues.apache.org/jira/browse/HIVE-7992
> Project: Hive
> Issue Type: Bug
> Components: Statistics
> Affects Versions: 0.13.1
> Reporter: Mostafa Mokhtar
> Assignee: Prasanth J
> Labels: performance
> Fix For: 0.14.0
>
>
> When StatsRulesProcFactory overflows it sets data size to 0 and as a result
> the Vertex will ask for a single task, this results in a fairly slow running
> query, most likely the overflow is a result of higher than usual number of
> rows.
> The class should detect an overflow and set a flag when an overflow occurs,
> if an overflow occurs StatsRulesProcFactory should request the maximum number
> of tasks for the vertex.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
