On Wed, Oct 7, 2015 at 9:06 PM, Kevin Burton <bur...@spinn3r.com> wrote:
> Let's say I have 10 nodes, I add 5 more, if I fail to run nodetool > cleanup, is excessive data transferred when I add the 6th node? IE do the > existing nodes send more data to the 6th node? > No. Streaming only streams ranges which are currently owned by the source and will be owned by the target. https://issues.apache.org/jira/browse/CASSANDRA-7764 Has some details on the type of edge cases one is exposed to if one does not run cleanup; mostly they involve moving a range away from a node and then back onto it. > but I also think that the excessive data will be removed in the next major > compaction and that nodetool cleanup just triggers a major compaction. > Nothing* removes data-which-doesn't-belong-on-the-node-it's-on but cleanup compaction (*or scrub). =Rob
