You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Duo Zhang (Jira)" <ji...@apache.org> on 2020/11/16 04:00:05 UTC
[jira] [Commented] (HBASE-25283) Undo meta replica specialization
inside of the region replica framework
[ https://issues.apache.org/jira/browse/HBASE-25283?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17232502#comment-17232502 ]
Duo Zhang commented on HBASE-25283:
-----------------------------------
Looking at the code, I think first we need to remove bunch of old (implicit) designs in the code. Such as how we retry, and the meaning of return true for ReplicationEndpoint.replicate method.
> Undo meta replica specialization inside of the region replica framework
> -----------------------------------------------------------------------
>
> Key: HBASE-25283
> URL: https://issues.apache.org/jira/browse/HBASE-25283
> Project: HBase
> Issue Type: Improvement
> Components: read replicas
> Reporter: Michael Stack
> Priority: Major
>
> Filing an issue to capture a [~zhangduo] suggestion made in PR that was then carried to a comment over in the design document attached to HBASE-18070 only it is deserving of its own issue.
> {quote}
> I do not like that we have a specialized implementation for meta replica inside the region replica framework. They are almost the same so they should share the same code base. I think this enhancement could also be applied to normal RegionReplicaReplicationEndpoint, what we need is the config for the max distance between different secondary replicas. For normal region replicas, the value will be small, which means we will soon block the replication if any replicas are slow, to save memory. For meta replicas, we could have a large default value to let the good replicas catch up fast.
> {quote}
>
> Lets use this issue to fill out more on what [~zhangduo] suggests above.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)