You're saying test coverage for single node operation? There is a little bit, just not a lot. But if we flip single node to RaftConsensus, I imagine we will immediately get a bunch of test coverage.
I think the upgrade path is pretty simple -- the only thing missing AFAIK is the last known RPC address of the single member in the cluster. I'm investigating whether that's something that we can fill in implicitly (don't see a reason why not, since it's the local process). Mike On Fri, May 27, 2016 at 8:28 PM, Todd Lipcon <[email protected]> wrote: > The idea is nice, but I also am not sure about the test coverage, etc. > Also, we need to make sure there is an upgrade path from LocalConsensus to > a single-node RaftConsensus. > > On Fri, May 27, 2016 at 7:59 PM, Mike Percy <[email protected]> wrote: > > > What ever was missing? Do you remember? > > > > Thanks, > > Mike > > > > > On May 27, 2016, at 6:18 PM, David Alves <[email protected]> > wrote: > > > > > > Hi Mike > > > > > > I'd be for it if we have the regular consensus implementation working > > (as in tested to perform reasonably) for single node. > > > Are we there, or close? > > > > > > Best > > > David > > > > > > On Fri, May 27, 2016 at 5:51 PM, Mike Percy < [email protected] [ > > [email protected]] > wrote: > > > I would like to delete LocalConsensus from Kudu before 1.0. Anyone > > opposed > > > to this? > > > > > > Thanks, > > > Mike > > > > > > -- > > > Mike Percy > > > Software Engineer, Cloudera > > > > > > -- > Todd Lipcon > Software Engineer, Cloudera >
