[jira] [Commented] (CASSANDRA-4358) Delay Joining the ring until bootstrap and its compaction completes

["Brandon Williams (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/CASSANDRA-4358?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13397498#comment-13397498 ] 

Brandon Williams commented on CASSANDRA-4358:
---------------------------------------------

Waiting for compaction to quiesce is tricky, because if the node continues to see more writes than it can handle, it will never happen.  This puts someone who is already over capacity in a situation where they can never add capacity to get out of it.

One thing we could do for LCS is include level information for the sstables, so at least the bootstrapping node won't do needless compactions.  This is straightforward when only streaming from one source, but admittedly I haven't fully thought out if merging from multiple sources would still be valid.
                
> Delay Joining the ring until bootstrap and its compaction completes
> -------------------------------------------------------------------
>
>                 Key: CASSANDRA-4358
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-4358
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Core
>    Affects Versions: 1.1.1, 1.2
>            Reporter: Vijay
>            Priority: Minor
>             Fix For: 1.2
>
>
> LCS adds a lot of small SST's while bootstrapping a node (after streaming), followed by a large amount of compactions. 
> We might want to do the following to speedup the compaction and reduce the impact to the incoming traffic:
> 1) Delay the node from joining the ring and wait until the node settles down with compactions.
> 2) Un-throttle the compactions and do it ASAP.
> 3) Once the node settles down we can enable thrift and join the ring.
> at the Least it might be better for us to avoid thrift traffic until the compaction completes.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira