You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@ignite.apache.org by "Ivan Pavlukhin (JIRA)" <ji...@apache.org> on 2018/10/23 12:24:00 UTC

[jira] [Commented] (IGNITE-9828) MVCC: Continuous query failover.

    [ https://issues.apache.org/jira/browse/IGNITE-9828?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16660559#comment-16660559 ] 

Ivan Pavlukhin commented on IGNITE-9828:
----------------------------------------

In ticket https://issues.apache.org/jira/browse/IGNITE-5935 partition counters "gap" closing was implemented. Gaps are closed in a way that all numeric counter values are consistent after PME. See method {{PartitionUpdateCounter.finalizeUpdateCountres}}. The missing part is CQ manager notification in situation of gap closing.

> MVCC: Continuous query failover.
> --------------------------------
>
>                 Key: IGNITE-9828
>                 URL: https://issues.apache.org/jira/browse/IGNITE-9828
>             Project: Ignite
>          Issue Type: Task
>          Components: mvcc
>            Reporter: Roman Kondakov
>            Priority: Major
>             Fix For: 2.8
>
>
> We need to implement failover procedure for CQ with MVCC caches. Thoughts:
> * Mvcc updates should be cached in the persistent off-heap data structures instead of heap.
> * CQ failover tests should be adopted for mvcc scenarios. See {{IgniteCacheQuerySelfTestSuite4}}
> * We need to ensure the correctness of CQ and partition counters consistency in cases when some of TX participants are in prepared state and others are marked as rollback only. It looks like it doesn't work properly now.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)