So I was able to get the schema agreeing on the two bad nodes, but I don't
particularly like the way that I did it. One at a time, I shut them down,
removed Schema* and Migration*, then copied over Schema* from another
working node. They then started up with the correct schema. Did I do
My understanding was that total is sum of all SSTables on disc even those not
being used currently and live is the sum of the SSTables on disc that are
being used currently.
Well I have a Cassandra 1.0.2 cluster in which all of the column families
are reporting live being greater than total. I
I figured it out. If I started node 2 first and used that as the seed for
the rest then the ring formed correctly. I guess the ring information
stored in the LocationInfo CF of node 1 was wrong.
For future information, when does this information get updated? All of the
nodes have different