[
https://issues.apache.org/jira/browse/IGNITE-10693?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16767218#comment-16767218
]
Igor Seliverstov edited comment on IGNITE-10693 at 2/13/19 2:07 PM:
--------------------------------------------------------------------
Seems we have wrong mapping logic in case of unstable topology. In the patch
reworked mapping is implemented.
the main flow looks so:
1) find all nodes that have all partitions for involved replicated caches
(optional step)
2) find nodes that have each partition of each involved partitioned cache.
Filter these nodes using previously collected nodes for replicated caches if
needed
3) reduce the nodes count for execution - use as little as possible nodes
having all needed partitions but not intersecting partitions between nodes.
was (Author: gvvinblade):
Seems we have wrong mapping logic in case of unstable topology. In the patch
reworked mapping implemented.
the main flow looks so:
1) find all nodes that have all partitions for involved replicated caches
(optional step)
2) find nodes that have each partition of each involved partitioned cache.
Filter these nodes using previously collected nodes for replicated caches if
needed
3) reduce the nodes count for execution - use as little as possible nodes
having all needed partitions but not intersecting partitions between nodes.
> MVCC TX: Random server restart tests became failed
> --------------------------------------------------
>
> Key: IGNITE-10693
> URL: https://issues.apache.org/jira/browse/IGNITE-10693
> Project: Ignite
> Issue Type: Bug
> Components: mvcc, sql
> Reporter: Igor Seliverstov
> Assignee: Igor Seliverstov
> Priority: Major
> Labels: failover, mvcc_stabilization_stage_1
> Fix For: 2.8
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> [one|https://ci.ignite.apache.org/project.html?projectId=IgniteTests24Java8&testNameId=7945125576049372508&branch=%3Cdefault%3E&tab=testDetails],
>
> [two|https://ci.ignite.apache.org/project.html?projectId=IgniteTests24Java8&testNameId=8412476034648229784&branch=%3Cdefault%3E&tab=testDetails],
>
> [three|https://ci.ignite.apache.org/project.html?projectId=IgniteTests24Java8&testNameId=254244004184327163&branch=%3Cdefault%3E&tab=testDetails],
> all these tests became failed after IGNITE-9630 has been merged to master.
> Seems there is an issue in partition calculating mechanism on unstable
> topology. I suppose the algorithm uses partitions on nodes just become
> primary while the partitions are in moving state.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
