[
https://issues.apache.org/jira/browse/IGNITE-8912?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16538971#comment-16538971
]
Vyacheslav Koptilin commented on IGNITE-8912:
---------------------------------------------
Hi [~pvinokurov]
It seems to me, your changes look correct.
At the moment when `GridDhtPartitionTopologyImpl.afterExchange()` is called,
the `exchange` process is already completed,
and so I can imagine one possible situation when we have a partition in the
`MOVING` state and there is no owner. Owner node left the cluster and
`discoCache` was updated by `discovery` thread, and as you mentioned, new
exchange will mark that partition as lost.
> PartitionLossPolicy.READ_ONLY_SAFE does not detect partition loss
> ------------------------------------------------------------------
>
> Key: IGNITE-8912
> URL: https://issues.apache.org/jira/browse/IGNITE-8912
> Project: Ignite
> Issue Type: Bug
> Components: cache
> Affects Versions: 2.5
> Reporter: Pavel Vinokurov
> Assignee: Pavel Vinokurov
> Priority: Major
> Attachments: MissedPartitionLostReproducer.java
>
>
> Cluster of 4 for a cache with 1 backup and READ_ONLY_SAFE .
> After forcefully killed two nodes, a partition lost without including in
> partitionsLost collection.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
