[
https://issues.apache.org/jira/browse/IGNITE-10897?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16761753#comment-16761753
]
Vladimir Ozerov commented on IGNITE-10897:
------------------------------------------
The issue is caused by "destroyCache" routine - it clears cache internals
before H2 tables are dropped. Correct fix is probably to have two-phase cache
destroy: on phase to prevent new operations on the cache, second phase is to
actually drop the cache when nobody holds any locks for sure.
This is complex thing. On hold for now.
> Blocked drop table operations cause strange issues
> --------------------------------------------------
>
> Key: IGNITE-10897
> URL: https://issues.apache.org/jira/browse/IGNITE-10897
> Project: Ignite
> Issue Type: Bug
> Components: sql
> Affects Versions: 2.6, 2.7
> Reporter: JiajunYang
> Priority: Major
>
> Reproduce steps to get a blocked drop table operation:
> 1.Create a table and put some data to a node with persistence enabled.
> 2.Do a select query targeting on the table which runs long time.
> 3.Drop the table before the select query end.
> Then u will see drop table operation blocks until the select query end.
> A strange issue cause by blocked drop table operations:
> 1.Do another drop table operation when there is a blocked drop table
> operation.This operation will also block.
> 2.Try to recreate the table with same name.Then you can see "table
> already exists" exception.
> 3.Try to drop the table again,then you can see
> "Table doesn't exist" exception.
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
