[
https://issues.apache.org/jira/browse/IGNITE-7039?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16630457#comment-16630457
]
Vladimir Ozerov edited comment on IGNITE-7039 at 9/27/18 2:41 PM:
------------------------------------------------------------------
This ticket appears to be related to IGNITE-6677, where we also parse
{{PreparedStatement}} in order to get the list of involved caches. In 6677 we
do that in more correct way - we use metadata of {{PreparedStatementEx}} in
order to re-use list of caches later.
For this reason it is better to wait for mentioned ticket to be resolved first,
and then re-use its cache ID calculation logic.
was (Author: vozerov):
This ticket appears to be related to IGNITE-6677, where we also parse
{{PreparedStatement}} in order to get the list of involved caches. However, in
this ticket we do that in more correct way - we use metadata of
{{PreparedStatementEx}} in order to re-use list of caches later.
For this reason it is better to wait for mentioned ticket to be resolved first,
and then re-use its cache ID calculation logic.
> SQL: local query should pin affected partitions
> -----------------------------------------------
>
> Key: IGNITE-7039
> URL: https://issues.apache.org/jira/browse/IGNITE-7039
> Project: Ignite
> Issue Type: Bug
> Components: sql
> Reporter: Vladimir Ozerov
> Assignee: Sergey Grimstad
> Priority: Major
> Labels: sql-stability
> Fix For: 2.8
>
> Attachments: 3194.patch
>
>
> When distributed query is executed, we pin cache partitions for particular
> topology version on map nodes [1]. However, it seems that we do no do that
> for local queries. This is a bug because partition with required data could
> be evicted from local node at any time, leading to incorrect results.
> [1]
> https://github.com/apache/ignite/blob/ignite-2.3/modules/indexing/src/main/java/org/apache/ignite/internal/processors/query/h2/twostep/GridMapQueryExecutor.java#L288
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
