[jira] [Commented] (IGNITE-13050) ClusterGroup that is recomputed on topology change

["Ilya Kasnacheev (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/IGNITE-13050?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17120811#comment-17120811 ] 

Ilya Kasnacheev commented on IGNITE-13050:
------------------------------------------

Thank you for this improvement, I have merged it to master!

> ClusterGroup that is recomputed on topology change
> --------------------------------------------------
>
>                 Key: IGNITE-13050
>                 URL: https://issues.apache.org/jira/browse/IGNITE-13050
>             Project: Ignite
>          Issue Type: Improvement
>          Components: compute
>            Reporter: Ivan Bessonov
>            Assignee: Ivan Bessonov
>            Priority: Major
>             Fix For: 2.9
>
>          Time Spent: 10m
>  Remaining Estimate: 0h
>
> Currently, ClusterGroup comes in two favors:
> One is a static set of UUIDs which will not change, second is predicate that is recomputed over ALL nodes on EVERY operation. This has bitten our client because recomputing of ClusterGroup happens in tcp-communication thread clogging it and delaying every operation in cluster. This is a major problem.
> It would be nice if there was a ClusterGroup with predicate which would recompute once per topology affinity change. Bonus points if it precisely tracks current topology with zero delay or overrun.
> Would be nice to upgrade firstNode/lastNode predicates to that mechanism since now they are static - topology changes but firstNode/lastNode projections don't, they may point to absent node.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)