[
https://issues.apache.org/jira/browse/CASSANDRA-8943?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14356042#comment-14356042
]
Yuki Morishita commented on CASSANDRA-8943:
-------------------------------------------
The problem with gossip available vnode that I first experimented is that
replica ranges change overtime during bootstrapping. Of course, we can
pre-calculate replica ranges before bootstrapping, though once we activate part
of vnodes, replication will be different from pre-calculated one. There still
can be missing or excess ranges in the middle of bootstrapping.
> Accept reads while bootstrapping for available ranges
> -----------------------------------------------------
>
> Key: CASSANDRA-8943
> URL: https://issues.apache.org/jira/browse/CASSANDRA-8943
> Project: Cassandra
> Issue Type: Sub-task
> Components: Core
> Reporter: Yuki Morishita
> Assignee: Yuki Morishita
> Priority: Minor
> Fix For: 3.0
>
>
> Last step for CASSANDRA-8494 is to make bootstrapping node accept read
> request and serve data for available keyspace/ranges.
> If data requested is not available, bootstrap node can either 1) send back
> 'Not Available' and let coordinator forward request to current replica node
> or 2) send proxy request to replica node and let it response to coordinator.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
