[
https://issues.apache.org/jira/browse/IGNITE-18044?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17646704#comment-17646704
]
Vladislav Pyatkov commented on IGNITE-18044:
--------------------------------------------
The root cause of the is that we do not recovery after restart in some case
when _TxCleanupCommand _was not handled in a RAFT sate machine
(_PartitionListener_). This command is read from log and try to apply on
_PartitionListener _after restart, but the collection of processing keys is
clear after restart. In this scenario, the command is applied and do nothing.
All write intents are till in the partition storage.
After the node started and a load steps into a some key of that transaction,
whose _TxCleanupCommand _did nothing, the node fails with the assertion error:
{noformat}Mismatched transaction id,
expectedTxId={000d780b-10c0-0000-face-5ad039564953},
actualTxId={000d780a-a0fd-0000-face-5ad039564953}{noformat}
There is possible recovery pending keys collection (_txsPendingRowIds_) using
scan command. But for this purpose required an implementation of issue
IGNITE-17467.
> ItIgniteNodeRestartTest#testTwoNodesRestartDirect is failed after schema
> recovery fix.
> --------------------------------------------------------------------------------------
>
> Key: IGNITE-18044
> URL: https://issues.apache.org/jira/browse/IGNITE-18044
> Project: Ignite
> Issue Type: Bug
> Reporter: Evgeny Stanilovsky
> Assignee: Vladislav Pyatkov
> Priority: Major
> Labels: ignite-3
>
> In [1] was introduced schema recovery fix, after the fix test [2] is failing
> with :
> {noformat}
> Caused by: java.lang.AssertionError: Mismatched transaction id,
> expectedTxId={000d780b-10c0-0000-face-5ad039564953},
> actualTxId={000d780a-a0fd-0000-face-5ad039564953}
> {noformat}
> [1] https://issues.apache.org/jira/browse/IGNITE-17986
> [2] ItIgniteNodeRestartTest#testTwoNodesRestartDirect
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
