On Sun, Jun 24, 2012 at 1:24 PM, Jonathan Ellis <jbel...@gmail.com> wrote: > Is it fair to say that 3881 and 4121 are the key tickets, and the others > are mostly dealing with fallout from that?
3881 just reduces the computational complexity of calculateNaturalEndpoints, which becomes (more of a )problem with many tokens per node. It doesn't introduce a functional change; It's pretty stand-alone 4121 introduces the ability to have more than one token per node, without doing anything with them. 4122 implements vnode-aware bootstrap, decommission, and node removal. It's also includes configuration for csv initial_tokens, and num_tokens. 4125 updates nodetool. 4127 is for the migrating an upgraded node from one to many tokens. > On Wed, Jun 20, 2012 at 4:25 PM, Eric Evans <eev...@acunu.com> wrote: >> So to followup on this, my take on the first discussion >> (http://www.mail-archive.com/dev@cassandra.apache.org/msg03837.html) >> was that there was some concern about how invasive this change would >> be. This was then followed by discussion about the trade-offs between >> trying to push such a big change in all at once, versus an incremental >> approach, and the associated risks if for some reason we weren't able >> to follow through. >> >> The point we're at now is our first shot at trying to thread this >> needle. As Sam says, it's at a point where it should be minimally >> useful, and should be backward compatible when run with 1 token per >> node (the default). The patches I am referring to again are linked >> from: >> >> https://issues.apache.org/jira/browse/CASSANDRA-4121 >> https://issues.apache.org/jira/browse/CASSANDRA-4122 >> https://issues.apache.org/jira/browse/CASSANDRA-4125 >> https://issues.apache.org/jira/browse/CASSANDRA-4127 >> >> What we most need now is feedback, and consensus from the project. >> Consensus on the proposed changes, the implementation, the scope of >> the first iteration, and subsequent steps. >> >> Thanks, -- Eric Evans Acunu | http://www.acunu.com | @acunu
