[
https://issues.apache.org/jira/browse/SOLR-6237?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14993781#comment-14993781
]
Mark Miller commented on SOLR-6237:
-----------------------------------
bq. Was also wondering what was the motivation behind this? Is this to make log
replays and recovery more robust?
* It takes non leader recovery and replica out of sync issues out of the
equation. There is no real recovery stage other than the leader replaying the
tlog.
* It removes the double replication you currently must have at the Solr and
HDFS level.
* Leader does not have to do work to distribute most updates to replicas.
* You can add some features that would be difficult or impossible to add
otherwise.
It's a good option with various tradeoffs, just like using autoAddReplicas with
HDFS or using the standard model.
> An option to have only leaders write and replicas read when using a shared
> file system with SolrCloud.
> ------------------------------------------------------------------------------------------------------
>
> Key: SOLR-6237
> URL: https://issues.apache.org/jira/browse/SOLR-6237
> Project: Solr
> Issue Type: New Feature
> Components: hdfs, SolrCloud
> Reporter: Mark Miller
> Assignee: Mark Miller
> Attachments: Unified Replication Design.pdf
>
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
