You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Vyacheslav Koptilin (Jira)" <ji...@apache.org> on 2022/09/23 13:55:00 UTC
[jira] [Updated] (IGNITE-10652) LocalWalModeNoChangeDuringRebalanceOnNonNodeAssignTest fails on TC
[ https://issues.apache.org/jira/browse/IGNITE-10652?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Vyacheslav Koptilin updated IGNITE-10652:
-----------------------------------------
Description:
LocalWalModeNoChangeDuringRebalanceOnNonNodeAssignTest fails on TC after splitting test into 3 for Atomic, Tx and Mvcc modes.
Seems, there is a race we were not able to detect before due to some reason.
The root cause of the issue is that the test assumes that the last joined server node will meet the following requirements:
- this node is affinity node
- affinity function will not choose this node as a primary node for any partition.
Unfortunately, the test does not provide any guarantee that the second requirement is met. So, the new server node (last joined server node) has one partition in moving state and disables the WAL on rebalance.
The obvious fix is providing a custom affinity functions that will guarantee that both of the requirements are met.
was:
LocalWalModeNoChangeDuringRebalanceOnNonNodeAssignTest fails on TC after splitting test into 3 for Atomic, Tx and Mvcc modes.
Seems, there is a race we were not able to detect before due to some reason.
> LocalWalModeNoChangeDuringRebalanceOnNonNodeAssignTest fails on TC
> ------------------------------------------------------------------
>
> Key: IGNITE-10652
> URL: https://issues.apache.org/jira/browse/IGNITE-10652
> Project: Ignite
> Issue Type: Bug
> Components: persistence
> Reporter: Andrey Mashenkov
> Assignee: Vyacheslav Koptilin
> Priority: Major
> Labels: MakeTeamcityGreenAgain
> Time Spent: 10m
> Remaining Estimate: 0h
>
> LocalWalModeNoChangeDuringRebalanceOnNonNodeAssignTest fails on TC after splitting test into 3 for Atomic, Tx and Mvcc modes.
> Seems, there is a race we were not able to detect before due to some reason.
> The root cause of the issue is that the test assumes that the last joined server node will meet the following requirements:
> - this node is affinity node
> - affinity function will not choose this node as a primary node for any partition.
> Unfortunately, the test does not provide any guarantee that the second requirement is met. So, the new server node (last joined server node) has one partition in moving state and disables the WAL on rebalance.
> The obvious fix is providing a custom affinity functions that will guarantee that both of the requirements are met.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)