[
https://issues.apache.org/jira/browse/IGNITE-18590?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Maksim Timonin updated IGNITE-18590:
------------------------------------
Description:
This is a bug introduced in the patch
[https://github.com/apache/ignite/pull/8822.]
When a transaction commits on primary node during TxRecovery procedure, it also
sends rollback request (GridDhtTxFinishRequest#commit = false) on backup nodes.
Then it's possible that such a request (commit = false) reaches a backup node
earlier than GridCacheTxRecoveryResponse(commit = true). Then primary node
will commit the transaction, and backup node rollbacks it. Then cluster will be
in inconsistent state.
> Transaction recovery tries rollback a tx concurrently with committing it
> ------------------------------------------------------------------------
>
> Key: IGNITE-18590
> URL: https://issues.apache.org/jira/browse/IGNITE-18590
> Project: Ignite
> Issue Type: Bug
> Reporter: Maksim Timonin
> Assignee: Maksim Timonin
> Priority: Major
> Fix For: 2.15
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> This is a bug introduced in the patch
> [https://github.com/apache/ignite/pull/8822.]
> When a transaction commits on primary node during TxRecovery procedure, it
> also sends rollback request (GridDhtTxFinishRequest#commit = false) on backup
> nodes.
> Then it's possible that such a request (commit = false) reaches a backup node
> earlier than GridCacheTxRecoveryResponse(commit = true). Then primary node
> will commit the transaction, and backup node rollbacks it. Then cluster will
> be in inconsistent state.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
