You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Roman Novichenok (JIRA)" <ji...@apache.org> on 2018/10/03 17:58:00 UTC

[jira] [Updated] (IGNITE-9775) IgniteCache.lostPartitions() is empty with single node remaining in the cluster

     [ https://issues.apache.org/jira/browse/IGNITE-9775?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Roman Novichenok updated IGNITE-9775:
-------------------------------------
    Summary: IgniteCache.lostPartitions() is empty with single node remaining in the cluster  (was: Cache.lostPartitions() is empty with single node remaining in the cluster)

> IgniteCache.lostPartitions() is empty with single node remaining in the cluster
> -------------------------------------------------------------------------------
>
>                 Key: IGNITE-9775
>                 URL: https://issues.apache.org/jira/browse/IGNITE-9775
>             Project: Ignite
>          Issue Type: Bug
>          Components: persistence
>    Affects Versions: 2.5, 2.6
>            Reporter: Roman Novichenok
>            Priority: Major
>
> When only one node remains in the cluster, cache.lostPartitions() is empty.
> Pull request: [https://github.com/novicr/ignite/pull/1] contains a test to show the issue.
> To reproduce:
>  # Turn on native persistence
>  # Create partitioned cache with 1 backup, sync replication, READ_ONLY_SAFE loss policy
>  # Start 4 nodes
>  # Persist enough data to make sure all nodes participate
>  # Stop 3 nodes
> Expected behavior:
>  * cache.lostPartitions().size() > 0
> Actual:
>  * cache.lostPartitions().size() == 0
>  
> Note, if more than one node is operational, the lostPartitions() will contain some number of partitions.  Don't know if the list it contains is complete.
>  
> The issue seems to be in the GridDhtPartitionTopologyImpl.detectLostPartitions()
> With a single node left in the cluster, the coordinator node is also the only one that contains any data, but localNode is skipped when examining node2Part maps to look for any partitions that should be marked LOST.
>  



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