Anton Gozhiy created DRILL-7022:
-----------------------------------
Summary: Partition pruning is not happening the first time after
the metadata auto refresh.
Key: DRILL-7022
URL: https://issues.apache.org/jira/browse/DRILL-7022
Project: Apache Drill
Issue Type: Bug
Affects Versions: 1.15.0
Reporter: Anton Gozhiy
*Data creation:*
# Create table:
{code:sql}
create table dfs.tmp.`orders`
partition by (o_orderstatus)
as select * from cp.`tpch/orders.parquet`
{code}
# Create table metadata:
{code:sql}
refresh table metadata dfs.tmp.`orders`
{code}
*Steps:*
# Modify the table to trigger metadata auto refresh:
{noformat}
hadoop fs -mkdir /tmp/orders/111
{noformat}
# Run the query:
{code:sql}
explain plan for
select * from dfs.tmp.`orders`
where o_orderstatus = 'O' and o_orderdate < '1995-03-10'
{code}
*Expected result:*
Partition pruning happens:
{noformat}
... numFiles=1, numRowGroups=1, usedMetadataFile=true ...
{noformat}
*Actual result:*
Partition pruning doesn't happen:
{noformat}
... numFiles=1, numRowGroups=3, usedMetadataFile=true
{noformat}
*Note:* It is being reproduced only the first time after auto refresh, after
repeating the query it works as expected.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
