[
https://issues.apache.org/jira/browse/CASSANDRA-9308?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14529067#comment-14529067
]
J.B. Langston edited comment on CASSANDRA-9308 at 5/5/15 7:03 PM:
------------------------------------------------------------------
3.0 is a long way off for a lot of production users though. And for the
scenario Brandon brings up, I think logging a warning for the user to run
repair may be preferable to forcing them to go through the whole bootstrap and
reindex process again, which in some cases can take days.
was (Author: jblangs...@datastax.com):
3.0 is a long way off for a lot of production users though.
> Decouple streaming from secondary index rebuild
> -----------------------------------------------
>
> Key: CASSANDRA-9308
> URL: https://issues.apache.org/jira/browse/CASSANDRA-9308
> Project: Cassandra
> Issue Type: Bug
> Reporter: J.B. Langston
>
> Currently, streaming is not considered complete until any secondary indexes
> on the table being streamed have been rebuilt. If any source replicas go
> down after streaming completes, but before the secondary indexes have been
> rebuilt, it will cause the bootstrap to fail, requiring the user to go
> through the whole bootstrap process again. Ideally, the two should be
> decoupled so that once the streaming is complete, the new node can complete
> the secondary index rebuild and successfully boostrap regardless of the
> status of the source replicas at that point.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
