[jira] [Updated] (IGNITE-11271) Investigate setting discardCustomMsgId to null in prepareNodeAddedMessage

["Maxim Muzafarov (Jira)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/IGNITE-11271?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Maxim Muzafarov updated IGNITE-11271:
-------------------------------------
    Fix Version/s:     (was: 2.8)

> Investigate setting discardCustomMsgId to null in prepareNodeAddedMessage
> -------------------------------------------------------------------------
>
>                 Key: IGNITE-11271
>                 URL: https://issues.apache.org/jira/browse/IGNITE-11271
>             Project: Ignite
>          Issue Type: Improvement
>            Reporter: Alexey Goncharuk
>            Priority: Major
>
> From debugging IGNITE-10935 it was discovered that NodeAddedMessage contains wrong state: pending messages are already filtered out by discard ID, but at the same time discardId and customDiscardId are set to non-null values. This resulted in a broken pending messages iterator on a newly added node: SkipIterator was skipping all pending messages until a valid discardId was received.
> The fix made in IGNITE-10935 was incomplete because we should have set both discardId and customDiscardId to null. However, after running TC tests it turned out that setting discardCustomMsgId to null resulted in duplicate custom events (the particular failed test is AuthenticationProcessorNodeRestartTest#testConcurrentAddUpdateRemoveNodeRestartServer)
> The reason behind the failed test is that some of the fired custom events are delivered multiple times. This should be investigated.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)