You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Anton Vinogradov (Jira)" <ji...@apache.org> on 2022/10/17 17:15:00 UTC
[jira] [Commented] (IGNITE-11611) If partition consistency cannot be restored during rebalance using counters the most recent partition data should be used.
[ https://issues.apache.org/jira/browse/IGNITE-11611?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17619011#comment-17619011 ]
Anton Vinogradov commented on IGNITE-11611:
-------------------------------------------
[~ascherbakov] , such approach may/will cause data loss.
Partition copies with not a "most recent" data may contain a data missed at "most recent" data owner.
Correct fix is to rebalance entries between LWM and HWM using the historical rebalance as well when possible (see. IGNITE-17738 for details)
When historihal rebalance is not possible we must fix the data using the recovery tool (see. IGNITE-15167 for details)
> If partition consistency cannot be restored during rebalance using counters the most recent partition data should be used.
> --------------------------------------------------------------------------------------------------------------------------
>
> Key: IGNITE-11611
> URL: https://issues.apache.org/jira/browse/IGNITE-11611
> Project: Ignite
> Issue Type: Improvement
> Reporter: Alexey Scherbakov
> Priority: Major
>
> This is possible if all owners are left, then returned and counters are inconsistent (no dedicated max counter node)
> In such a case the "most recent" partition can be used for supplying purposes.
> We should track topology history to find the node with "most recent" updates.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)