[
https://issues.apache.org/jira/browse/IGNITE-3406?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15395170#comment-15395170
]
Dmitry Karachentsev commented on IGNITE-3406:
---------------------------------------------
Able to reproduce only first case, only for interceptor, and only for
TRANSACTIONAL cache. Fix requires deeper investigation of how tx cache works.
> Interceptor and continuous query can get incorrect old value during
> rebalancing
> -------------------------------------------------------------------------------
>
> Key: IGNITE-3406
> URL: https://issues.apache.org/jira/browse/IGNITE-3406
> Project: Ignite
> Issue Type: Bug
> Components: cache
> Affects Versions: 1.6
> Reporter: Valentin Kulichenko
> Assignee: Dmitry Karachentsev
> Fix For: 1.7
>
>
> During the rebalancing the old value passed to interceptor or continuous
> query can't be wrong due to scenarios described below.
> Scenario 1.
> # New node joins.
> # Update request arrives to the new node before the entry for the same key is
> rebalanced to this node.
> # Listeners are notified with {{null}} old value.
> In this case we should use {{GridDhtForceKeysRequest}} to get the old value.
> Scenario 2.
> # Entry is updated on primary node.
> # New backup node joins and the entry with already updated value is
> rebalanced from the primary node.
> # Update request arrives to this new backup node.
> # Listeners are notified with the old value which is the same as new value.
> In this case we should send the old value with the update request in case
> there is ongoing rebalancing and cache has interceptor configured or
> continuous queries registered.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
