[
https://issues.apache.org/jira/browse/IGNITE-22245?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Alexander Lapin updated IGNITE-22245:
-------------------------------------
Ignite Flags: (was: Docs Required,Release Notes Required)
> Write intents are not switched if primary changes
> -------------------------------------------------
>
> Key: IGNITE-22245
> URL: https://issues.apache.org/jira/browse/IGNITE-22245
> Project: Ignite
> Issue Type: Bug
> Reporter: Kirill Sizov
> Assignee: Kirill Sizov
> Priority: Major
> Fix For: 3.0.0-beta2
>
> Time Spent: 50m
> Remaining Estimate: 0h
>
> If during the cleanup process the cleanup request fails to execute on an
> initially enlisted node the durable cleanup logic retries the cleanup request
> on the new partition primaries.
> Since the new primary node does not have in-memory transactional state, it
> skips write intent switch and reports successful cleanup.
> As a result we get a successfully completed transaction with write intents
> instead of real data in the storage. We had been covered with write intent
> resolution before vacuum was introduced.
> Now, if the state has been vacuumized, there is no chance to correctly
> resolve such write intents and we get corrupted data.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
