You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Bharath Vissapragada (Jira)" <ji...@apache.org> on 2020/07/01 23:03:00 UTC
[jira] [Comment Edited] (HBASE-24666) Offload the replication
source/sink job to independent Replication Server
[ https://issues.apache.org/jira/browse/HBASE-24666?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17149734#comment-17149734 ]
Bharath Vissapragada edited comment on HBASE-24666 at 7/1/20, 11:02 PM:
------------------------------------------------------------------------
I like this idea of de-coupling replication subsystem from the region server. We are brainstorming similar ideas at my employer for the very same reasons of write amplification to multiple peers and we have plans to run custom replication endpoints (with 3rd party code and external dependencies). This decoupling provides a clean way to implement all of these. I think this is also a good opportunity to clean up the nexus of Replication* classes. Added my thoughts to the design doc, thanks for putting it together.
was (Author: bharathv):
I like this idea of de-coupling replication subsystem from the region server. We are brainstorming similar ideas at my employer for the very same reasons of write amplification to multiple peers and we have plans to run customer replication endpoints (with 3rd party code and external dependencies). This decoupling provides a clean way to implement all of these. I think this is also a good opportunity to clean up the nexus of Replication* classes. Added my thoughts to the design doc, thanks for putting it together.
> Offload the replication source/sink job to independent Replication Server
> -------------------------------------------------------------------------
>
> Key: HBASE-24666
> URL: https://issues.apache.org/jira/browse/HBASE-24666
> Project: HBase
> Issue Type: Umbrella
> Reporter: Guanghao Zhang
> Priority: Major
>
> The basic idea is add a role "ReplicationServer" to take the replication source/sink job. HMaster is responsible for scheduling the replication job to different ReplicationServer.
> [link Design doc|https://docs.google.com/document/d/16kRPVGctFSf__nC3yaVZmAm3GTxIbHefekKC_rMmTw8/edit?usp=sharing]
> Suggestions are welcomed. Thanks.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)