The issue has something to do with having -ROOT- and .META. on the same server which is a rare event given that the assignment function will only do this if if finds the cluster is made of one server only (On ec2, where the issue was found, using our ec2 scripts, this is a more likely given how there can be a lag as servers come in after master startup).
I have been trying to reproduce and have not been having much luck. I'm sure there is something in here and with more time, we'll narrow in on the trigger but my thought, after talking with Andrew offline, is to change his -1 to a +0, that this should not sink the current release candidate, and that he we'll get it next time around in 0.20.6. St.Ack On Sat, Jun 19, 2010 at 1:48 PM, Andrew Purtell <apurt...@apache.org> wrote: > See my other mail today. > > If I have it wrong, please feel free to change my vote to 0. > > I will be out of touch for 24 hours+. > > - Andy > > > > > >
