[ 
https://issues.apache.org/jira/browse/IGNITE-18261?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Roman Puchkovskiy updated IGNITE-18261:
---------------------------------------
    Description: 
For IGNITE-18085, we need a mechanism for a distributed, guaranteed, in-order 
delivery of events concerning the Logical Topology to all nodes of the cluster. 
CMG nodes get the corresponding events automatically thanks to the RAFT 
replication mechanism. To extend delivery (using RAFT) to the whole cluster we 
can use learners: run a learner per non-CMG node.

When a CMG raft group is started, learners set needs to be computed as all 
nodes of the cluster that are not in the CMG. When the logical topology gets 
changed (or when CMG member set gets changed), we need to recalculate learners 
set and use {{changePeers}} accordingly.

  was:For IGNITE-18085, we need a mechanism for a distributed, guaranteed, 
in-order delivery of events concerning the Logical Topology to all nodes of the 
cluster. CMG nodes get the corresponding events automatically thanks to the 
RAFT replication mechanism. To extend delivery (using RAFT) to the whole 
cluster we can use learners: run a learner per non-CMG node.


> Use RAFT learners to get updates of logical topology on non-CMG nodes
> ---------------------------------------------------------------------
>
>                 Key: IGNITE-18261
>                 URL: https://issues.apache.org/jira/browse/IGNITE-18261
>             Project: Ignite
>          Issue Type: Improvement
>            Reporter: Roman Puchkovskiy
>            Assignee: Roman Puchkovskiy
>            Priority: Major
>              Labels: ignite-3
>             Fix For: 3.0.0-beta2
>
>
> For IGNITE-18085, we need a mechanism for a distributed, guaranteed, in-order 
> delivery of events concerning the Logical Topology to all nodes of the 
> cluster. CMG nodes get the corresponding events automatically thanks to the 
> RAFT replication mechanism. To extend delivery (using RAFT) to the whole 
> cluster we can use learners: run a learner per non-CMG node.
> When a CMG raft group is started, learners set needs to be computed as all 
> nodes of the cluster that are not in the CMG. When the logical topology gets 
> changed (or when CMG member set gets changed), we need to recalculate 
> learners set and use {{changePeers}} accordingly.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to