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 Kondakov (JIRA)" <ji...@apache.org> on 2019/01/31 16:24:00 UTC

[jira] [Commented] (IGNITE-10261) MVCC: cache operation may hang during late affinity assignment.

    [ https://issues.apache.org/jira/browse/IGNITE-10261?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16757438#comment-16757438 ] 

Roman Kondakov commented on IGNITE-10261:
-----------------------------------------

The root cause of the bug was the case when get request is mapped to the node with partition which is being preloaded (partition is not in the {{OWNING}} state), this causes a {{ForceKeyRequest}} to the supplier. This request is not supported by MVCC caches, even more, there are intentions to get rid of it for non-MVCC caches too (IGNITE-10251).
So the decision is to prohibit {{ForceKeyRequest}} for MVCC caches and send error back to the near node if partition is not in the {{OWNING}} state. In this case near node resends request to the next affinity node to obtain a value, until the {{OWNING}} partition is found.

> MVCC: cache operation may hang during late affinity assignment.
> ---------------------------------------------------------------
>
>                 Key: IGNITE-10261
>                 URL: https://issues.apache.org/jira/browse/IGNITE-10261
>             Project: Ignite
>          Issue Type: Bug
>          Components: mvcc
>    Affects Versions: 2.7
>            Reporter: Andrew Mashenkov
>            Assignee: Roman Kondakov
>            Priority: Critical
>              Labels: failover, mvcc_stabilization_stage_1
>             Fix For: 2.8
>
>          Time Spent: 10m
>  Remaining Estimate: 0h
>
> ForceKey response processing fails with ClassCastException with Mvcc mode that causes test hanging.
> Issue can be easily reproduced with backups > 0 and disabled rebalance. See GridCacheDhtPreloadPutGetSelfTest.testPutGetNone1().
> Also CacheLateAffinityAssignmentTest.testRandomOperations() hangs sometimes due to same reason.
>  
> {noformat}
> java.lang.ClassCastException: org.apache.ignite.internal.processors.cache.distributed.dht.GridDhtCacheEntry cannot be cast to org.apache.ignite.internal.processors.cache.mvcc.MvccVersionAware
>  at org.apache.ignite.internal.processors.cache.distributed.dht.preloader.GridDhtForceKeysFuture$MiniFuture.onResult(GridDhtForceKeysFuture.java:545)
>  at org.apache.ignite.internal.processors.cache.distributed.dht.preloader.GridDhtForceKeysFuture.onResult(GridDhtForceKeysFuture.java:202)
>  at org.apache.ignite.internal.processors.cache.distributed.dht.GridDhtCacheAdapter.processForceKeyResponse(GridDhtCacheAdapter.java:180)
>  at org.apache.ignite.internal.processors.cache.distributed.dht.GridDhtTransactionalCacheAdapter$11.onMessage(GridDhtTransactionalCacheAdapter.java:208)
>  at org.apache.ignite.internal.processors.cache.distributed.dht.GridDhtTransactionalCacheAdapter$11.onMessage(GridDhtTransactionalCacheAdapter.java:206)
>  at org.apache.ignite.internal.processors.cache.distributed.dht.GridDhtCacheAdapter$MessageHandler.apply(GridDhtCacheAdapter.java:1434)
>  at org.apache.ignite.internal.processors.cache.distributed.dht.GridDhtCacheAdapter$MessageHandler.apply(GridDhtCacheAdapter.java:1416)
>  at org.apache.ignite.internal.processors.cache.GridCacheIoManager.processMessage(GridCacheIoManager.java:1054)
> {noformat}
>  



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