[
https://issues.apache.org/jira/browse/IGNITE-9769?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16656725#comment-16656725
]
Ilya Lantukh commented on IGNITE-9769:
--------------------------------------
{noformat}if(affNodesCnt!=ownerNodesCnt||!affNodes.containsAll(owners)||
(waitEvicts&&loc!=null&&loc.state()!=GridDhtPartitionState.OWNING)){
if(i%50==0)
LT.warn(log(),"Waiting for topology map update ["+
"igniteInstanceName="+g.name()+
", cache="+cfg.getName()+
", cacheId="+dht.context().cacheId()+
", topVer="+top.readyTopologyVersion()+
", p="+p+
", affNodesCnt="+affNodesCnt+
", ownersCnt="+ownerNodesCnt+
", affNodes="+F.nodeIds(affNodes)+
", owners="+F.nodeIds(owners)+
", topFut="+topFut+
", locNode="+g.cluster().localNode()+']');
{noformat}
Looks like we need to check not only that affNodes.containsAll(owners), but
also that order of nodes in these collections (or at least first nodes, that
are primaries) is the same.
> IgniteCacheAtomicProtocolTest.testPutReaderUpdate1 is flaky
> -----------------------------------------------------------
>
> Key: IGNITE-9769
> URL: https://issues.apache.org/jira/browse/IGNITE-9769
> Project: Ignite
> Issue Type: Task
> Reporter: Ryabov Dmitrii
> Assignee: Ryabov Dmitrii
> Priority: Trivial
> Labels: MakeTeamcityGreenAgain
> Fix For: 2.8
>
>
> {{IgniteCacheAtomicProtocolTest.testPutReaderUpdate1}} and
> {{IgniteCacheAtomicProtocolTest.testPutReaderUpdate2}} are flaky.
> In the {{#readerUpdateDhtFails}} method we blocks
> {{GridDhtAtomicNearResponse}} messages and do put operation. Put should hangs
> always, but sometimes it doesn't.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
