[
https://issues.apache.org/jira/browse/IGNITE-10104?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16774010#comment-16774010
]
Roman Kondakov commented on IGNITE-10104:
-----------------------------------------
The main idea of the changes is do not lock keys during the query execution, as
it was before, but instead, obtain an iterator from the query on the near node
and stream keys from this iterator to the primary nodes with {{LOCK}}
operation. Exactly as it is works in DML queries with reducer.
> MVCC TX: client SFU doesn't work on replicated caches
> -----------------------------------------------------
>
> Key: IGNITE-10104
> URL: https://issues.apache.org/jira/browse/IGNITE-10104
> Project: Ignite
> Issue Type: Bug
> Components: mvcc, sql
> Reporter: Igor Seliverstov
> Assignee: Roman Kondakov
> Priority: Major
> Labels: mvcc_stabilization_stage_1, transactions
> Fix For: 2.8
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> When select for update executes from client node the execution is sent to
> random owning node. On that node dht enlist operation is started what causes
> an assertion error because dht enlist operation implies that the node is
> primary for all processed keys.
> see
> {{CacheMvccReplicatedBackupsTest.testBackupsCoherenceWithLargeOperations}}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
