[
https://issues.apache.org/jira/browse/IGNITE-13572?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17217589#comment-17217589
]
Alexey Goncharuk commented on IGNITE-13572:
-------------------------------------------
[~ivandasch], looks like the issue affects only caches with 0 backups because
we have an invalid optimization in {{IndexingQueryFilterImpl#forCache}}:
{code}
// No backups and explicit partitions -> nothing to filter.
if (cache.configuration().getBackups() == 0 && parts == null)
return null;
{code}
As a result, we skip filtering for such caches and allow query to read from
RENTING partitions. Removing this code fixes the test.
> Duplicates in select query during partition eviction.
> -----------------------------------------------------
>
> Key: IGNITE-13572
> URL: https://issues.apache.org/jira/browse/IGNITE-13572
> Project: Ignite
> Issue Type: Bug
> Components: sql
> Affects Versions: 2.9, 2.8.1
> Reporter: Ivan Daschinskiy
> Priority: Critical
> Attachments: SqlPartitionEvictionTest.java
>
>
> Scenario:
> # Start 2 nodes with indexed atomic partitioned cache with 0 backups.
> # Load sufficient amout of data (or emulate slow removal from idx)
> # Start another node.
> # Perform SELECT * FROM <TABLE>.
> Query result contains duplicates, result size is significantly bigger than
> expected cache size.
> Reproducer is attached.
> Reproduced on 2.8.1, ongoing 2.9 and master
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
