[
https://issues.apache.org/jira/browse/HDDS-11805?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ivan Andika updated HDDS-11805:
-------------------------------
Description:
-Currently ReplicationManager start and stop command is only processed by the
leader which will either start or stop the relevant threads as well as set the
flag "running". However, if there is a leader change, the new SCM leader will
not see the previous start / stop subcommand.-
-We can replicate this start and stop command to the followers so that each SCM
has a consistent view of whether replication should run.- (This is wrong,
replicationmanager start and stop command would broadcast to all SCM nodes.
Similar to ContainerBalancer, we can make ReplicationManager implements
StatefulService.
was:
Currently ReplicationManager start and stop command is only processed by the
leader which will either start or stop the relevant threads as well as set the
flag "running". However, if there is a leader change, the new SCM leader will
not see the previous start / stop subcommand.
We can replicate this start and stop command to the followers so that each SCM
has a consistent view of whether replication should run.
Similar to ContainerBalancer, we can make ReplicationManager implements
StatefulService.
> Replicate the ReplicationManager start and stop command
> -------------------------------------------------------
>
> Key: HDDS-11805
> URL: https://issues.apache.org/jira/browse/HDDS-11805
> Project: Apache Ozone
> Issue Type: Sub-task
> Reporter: Ivan Andika
> Assignee: Peter Lee
> Priority: Major
> Labels: pull-request-available
>
> -Currently ReplicationManager start and stop command is only processed by the
> leader which will either start or stop the relevant threads as well as set
> the flag "running". However, if there is a leader change, the new SCM leader
> will not see the previous start / stop subcommand.-
> -We can replicate this start and stop command to the followers so that each
> SCM has a consistent view of whether replication should run.- (This is wrong,
> replicationmanager start and stop command would broadcast to all SCM nodes.
> Similar to ContainerBalancer, we can make ReplicationManager implements
> StatefulService.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
