[
https://issues.apache.org/jira/browse/HIVE-21817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16855327#comment-16855327
]
Gopal V commented on HIVE-21817:
--------------------------------
Do you have a FIL following a TS?
The simplest pattern looks like
{code}
select count(1) from calls c where year_partition=2018 and date_time_date
between DATE'2018-01-01' and DATE'2018-02-01';
{code}
My benchmark has 86 thousand queries, not every query is hitting this - so I'll
try to narrow down the queries which hit this, but the specific call is coming
from the PCR which detects a TS+FIL
I have a generic call-trace, which goes backwards from
https://github.com/apache/hive/blob/master/ql/src/java/org/apache/hadoop/hive/ql/optimizer/pcr/PcrOpProcFactory.java#L110
+
https://github.com/apache/hive/blob/master/ql/src/java/org/apache/hadoop/hive/ql/optimizer/Optimizer.java#L124
!pcr-cache-miss.png!
> CBO: PartitionPruner cache-miss from Calcite planner
> ----------------------------------------------------
>
> Key: HIVE-21817
> URL: https://issues.apache.org/jira/browse/HIVE-21817
> Project: Hive
> Issue Type: Bug
> Components: CBO
> Reporter: Gopal V
> Assignee: Jesus Camacho Rodriguez
> Priority: Major
> Attachments: calcite-ppr.png, pcr-cache-miss.png
>
>
> !calcite-ppr.png!
> {code}
> public static PrunedPartitionList prune(Table tab, ExprNodeDesc prunerExpr,
> HiveConf conf, String alias, Map<String, PrunedPartitionList>
> prunedPartitionsMap)
> throws SemanticException {
> vs
> public PrunedPartitionList getPrunedPartitions(String alias,
> TableScanOperator ts)
> throws SemanticException {
> PrunedPartitionList partsList = opToPartList.get(ts);
> {code}
> Only the 2nd one hits the PPR cache.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
