You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Aleksandr Chesnokov (Jira)" <ji...@apache.org> on 2026/05/05 12:13:00 UTC
[jira] [Updated] (IGNITE-28280) Investigate disabling (fully or partially) historical rebalancing during Rolling Upgrade process
[ https://issues.apache.org/jira/browse/IGNITE-28280?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Aleksandr Chesnokov updated IGNITE-28280:
-----------------------------------------
Summary: Investigate disabling (fully or partially) historical rebalancing during Rolling Upgrade process (was: Disable (fully or partially) historical rebalancing during Rolling Upgrade process)
> Investigate disabling (fully or partially) historical rebalancing during Rolling Upgrade process
> ------------------------------------------------------------------------------------------------
>
> Key: IGNITE-28280
> URL: https://issues.apache.org/jira/browse/IGNITE-28280
> Project: Ignite
> Issue Type: Improvement
> Reporter: Sergey Chugunov
> Assignee: Aleksandr Chesnokov
> Priority: Major
> Labels: IEP-132, ise
> Fix For: 2.19
>
>
> Historical rebalancing involves sending WAL segments from the nodes with up-to-date data to the outdated nodes.
> But WAL segments are only guaranteed to be backwards compatible. Because of that a situation is possiblewhen an old node leaves and then returns back to the topology and a new node starts sending WAL segments to it. If format of WAL segments has changed between versions the old node will fail on reading the segments.
> To avoid this we need to make sure that new nodes never use historical rebalance to update old nodes. It could be done by a total prohibition of historical rebalancing during RU or by falling back to full rebalancing only in case of new nodes sending data to old nodes.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)