[
https://issues.apache.org/jira/browse/RATIS-1592?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17554860#comment-17554860
]
Yaolong Liu commented on RATIS-1592:
------------------------------------
[~William Song]
Thanks for the reply, I konw the scenario, but still have some questions:
If in the above scenario:
1. D sends setConfigurationCAS(old=[A,B], new=[A,B,D])
2. E sends setConfigurationCAS(old=[A,B], new=[A,B,E])
Let's assume that E is successful for the first time. At this time, the cluster
is A B E, and then a new round of changes will occur immediately to process the
request issued by D, and finally the cluster will become A B D. Therefore, if
we only want to ensure strict 3 copies, the current code is also satisfied, but
with an additional round of elections. The scenario described above does exist
in add-mode, can we consider CAS-mode a sub-mode of add mode?
> Concurrent setConfiguration
> ---------------------------
>
> Key: RATIS-1592
> URL: https://issues.apache.org/jira/browse/RATIS-1592
> Project: Ratis
> Issue Type: Improvement
> Reporter: Song Ziyang
> Priority: Major
>
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
