On Mon, Jul 14, 2014 at 11:12 AM, Denning, Michael < michael.denn...@kavokerrgroup.com> wrote:
> 3 node cluster is in production. It’d difficult for me to get sign off > on the change control to upgrade it. The 6 node cluster is already stood > up (in aws). In an ideal scenario I’d just be able to bring the data over > to the new cluster. > Ok, use the "copy the sstables" method from the previous link? 1) fork writes so all writes go to both clusters 2) nodetool flush on source cluster 3) copy all sstables to all target nodes, being careful to avoid name collision (use rolling restart, probably, "refresh" is unsafe) 4) run cleanup on target nodes (this will have the same effect as doing an upgradesstables, as a bonus) 5) turn off writes to old cluster/turn on reads to new cluster If I were you, I would strongly consider not using vnodes on your new cluster. Unless you are very confident the cluster will grow above appx 10 nodes in the near future, you are likely to Just Lose from vnodes. =Rob
