[jira] [Commented] (CASSANDRA-10287) nodetool rebuild does not work with join_ring=false

["Yuki Morishita (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/CASSANDRA-10287?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15386571#comment-15386571 ] 

Yuki Morishita commented on CASSANDRA-10287:
--------------------------------------------

Current rebuild assumes that the rebuilding node be bootstrapped. That is, the nodes in the cluster know the rebuilding node is a part of the cluster and its token ranges. Once bootstrapped, then rebuild with join_ring=false works.

We may be able to update rebuild by including bootstrap tokens and stream data regarding pending ranges, but current 'join_ring=false' state seems fragile. Bootstrap tokens are not persisted when in 'join_ring=false_but_not_bootstrapped' state, so when the node restarted before joining the ring, bootstrap tokens can change espeicially with vnode random token generation. I think what we need is to define more concrete definitions of the node states.

> nodetool rebuild does not work with join_ring=false
> ---------------------------------------------------
>
>                 Key: CASSANDRA-10287
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-10287
>             Project: Cassandra
>          Issue Type: Bug
>            Reporter: Tom van den Berge
>            Assignee: Paulo Motta
>
> I'm setting up a new data center as described in http://docs.datastax.com/en/cassandra/2.0/cassandra/operations/ops_add_dc_to_cluster_t.html, with one change: I'm starting the new node with join_ring=false, because I want to prevent reads to be routed to the new node.
> When starting nodetool rebuild as described in 7b, the command immediately returns. In the log file, I can see a message that states the the streaming is started. Then nothing happens, not even after a few hours of waiting.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)