[
https://issues.apache.org/jira/browse/IGNITE-19471?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Kirill Gusakov updated IGNITE-19471:
------------------------------------
Description:
Algorithm:
* TableManager receives the pending keys update inside the pending keys
listener: TableManager check if the current node is a leader of partition and
send the RebalanceRequest to PrimaryReplica through ReplicaService (TODO: do we
have an ability to send the message to PrimaryReplica directly?)
* PrimaryReplica receives this message and initiate the changePeers command to
the leader of RAFT group (TODO: should this call be protected by the leaderTerm
as the current call inside the TableManager?)
Other scenarios:
* onLeadeElected event RebalanceRaftGroupEventsListener must send the
RebalanceRequest message to PrimaryReplica instead of direct call of
changePeersAsync
* onReconfigurationError event listener RebalanceRaftGroupEventsListener must
send the RebalanceRequest also.
was:
Algorithm:
* TableManager receives the pending keys update inside the pending keys
listener: TableManager check if the current node is a leader of partition and
send the RebalanceRequest to PrimaryReplica throight ReplicaService (TODO: do
we have an ability to send the message to PrimaryReplica directly?)
* PrimaryReplica receives this message and initiate the changePeers command to
the leader of RAFT group (TODO: should this call be protected by the leaderTerm
as the current call inside the TableManager?)
Other scenarios:
* onLeadeElected event RebalanceRaftGroupEventsListener must send the
RebalanceRequest message to PrimaryReplica instead of direct call of
changePeersAsync
* onReconfigurationError event listener RebalanceRaftGroupEventsListener must
send the RebalanceRequest also.
> Move changePeers calls logic to PrimaryReplica
> ----------------------------------------------
>
> Key: IGNITE-19471
> URL: https://issues.apache.org/jira/browse/IGNITE-19471
> Project: Ignite
> Issue Type: Task
> Reporter: Kirill Gusakov
> Priority: Major
>
> Algorithm:
> * TableManager receives the pending keys update inside the pending keys
> listener: TableManager check if the current node is a leader of partition and
> send the RebalanceRequest to PrimaryReplica through ReplicaService (TODO: do
> we have an ability to send the message to PrimaryReplica directly?)
> * PrimaryReplica receives this message and initiate the changePeers command
> to the leader of RAFT group (TODO: should this call be protected by the
> leaderTerm as the current call inside the TableManager?)
> Other scenarios:
> * onLeadeElected event RebalanceRaftGroupEventsListener must send the
> RebalanceRequest message to PrimaryReplica instead of direct call of
> changePeersAsync
> * onReconfigurationError event listener RebalanceRaftGroupEventsListener
> must send the RebalanceRequest also.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
