rack aware basically means it has some knowledge of cluster topology so it can (a) do reads from the lowest-latency replica, and (b) distribute replicas across the ring better, e.g. at least one copy in each datacenter. other than that there is nothing special about the rack or data center as a unit of machines, so the answer is, the same thing as when you lose a single machine, only repeated. :)
On Wed, Nov 25, 2009 at 1:06 PM, Freeman, Tim <tim.free...@hp.com> wrote: > I suppose the point behind RackAware is that you can lose one rack or > datacenter and then recover somehow. How would the recovery work? I'd > probably have part of my cluster left, the replication factor would be three, > and I'd have less than a quorum for some of the rows. I'd have to add a > bunch of new nodes in the remaining datacenter, then I'd have to somehow > initiate a read repair to get a quorum back. I can't fill in the details. > Is there a document I should read about this? > > Tim Freeman > Email: tim.free...@hp.com > Desk in Palo Alto: (650) 857-2581 > Home: (408) 774-1298 > Cell: (408) 348-7536 (No reception business hours Monday, Tuesday, and > Thursday; call my desk instead.) > > >
