[ 
https://issues.apache.org/jira/browse/IGNITE-20701?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Pavel Tupitsyn updated IGNITE-20701:
------------------------------------
    Description: 
h2. Problem

* *Partition Leader* and *Primary Partition Replica* can be two different 
nodes. 
* Primary replica can handle requests more efficiently
* *InternalTableImpl#assignments* is used for client partition awareness, but 
it returns leaders

h2. Action Items

* Fix *ClientTablePartitionAssignmentGetRequest* to return primary replica per 
partition
* Fix *ClientInboundMessageHandler.onPartitionAssignmentChanged* to notify 
about primary change (lease based?)

All the fixes are server-side; in theory, we don't need to fix the clients.

Consult [~sanpwc] for more details.

  was:
h2. Problem

* *Partition Leader* and *Primary Partition Replica* can be two different 
nodes. 
* Primary replica can handle requests more efficiently
* *InternalTableImpl#assignments* is used for client partition awareness, but 
it returns leaders

h2. Action Items

* Fix *ClientTablePartitionAssignmentGetRequest* to return primary replica per 
partition
* Fix *ClientInboundMessageHandler.onPartitionAssignmentChanged* to notify 
about primary change (lease based?)
* Fix *IgniteCompute.executeColocated* the same way

All the fixes are server-side; in theory, we don't need to fix the clients.

Consult [~sanpwc] for more details.


> Thin 3.0: Partition awareness uses leader instead of primary replica
> --------------------------------------------------------------------
>
>                 Key: IGNITE-20701
>                 URL: https://issues.apache.org/jira/browse/IGNITE-20701
>             Project: Ignite
>          Issue Type: Bug
>          Components: thin client
>    Affects Versions: 3.0.0-beta1
>            Reporter: Pavel Tupitsyn
>            Assignee: Pavel Tupitsyn
>            Priority: Critical
>              Labels: ignite-3
>             Fix For: 3.0.0-beta2
>
>          Time Spent: 10m
>  Remaining Estimate: 0h
>
> h2. Problem
> * *Partition Leader* and *Primary Partition Replica* can be two different 
> nodes. 
> * Primary replica can handle requests more efficiently
> * *InternalTableImpl#assignments* is used for client partition awareness, but 
> it returns leaders
> h2. Action Items
> * Fix *ClientTablePartitionAssignmentGetRequest* to return primary replica 
> per partition
> * Fix *ClientInboundMessageHandler.onPartitionAssignmentChanged* to notify 
> about primary change (lease based?)
> All the fixes are server-side; in theory, we don't need to fix the clients.
> Consult [~sanpwc] for more details.



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

Reply via email to