[jira] [Commented] (IGNITE-10043) Lost partitions list is reset when only one server is alive in the cluster

["Stanislav Lukyanov (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/IGNITE-10043?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16667073#comment-16667073 ] 

Stanislav Lukyanov commented on IGNITE-10043:
---------------------------------------------

The issue is caused by the fact that with BLT lost partitions stay on the offline nodes.
It looks like GridDhtPartitionTopologyImpl only marks partitions as LOST on non-local online servers. When we have no non-local online servers, we don't have a place to have a LOST mark. See GridDhtPartitionTopologyImpl::detectLostPartition.

> Lost partitions list is reset when only one server is alive in the cluster
> --------------------------------------------------------------------------
>
>                 Key: IGNITE-10043
>                 URL: https://issues.apache.org/jira/browse/IGNITE-10043
>             Project: Ignite
>          Issue Type: Bug
>            Reporter: Stanislav Lukyanov
>            Priority: Major
>
> When server nodes are killed one by one IgniteCache.lostPartitions() returns non-empty list up until there is just one node left. When only one node is alive, the lostPartitions() returns an empty collection even though the partition loss events has been fired and the partitions are missing.
> This only affects cluster with persistence enabled.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)