Ha, I figured out what was going on. I was totally ignoring a test index
that had been balanced onto two of the nodes and
so the 3 indexes' shards were balanced across ALL the nodes, my assumption
is that each index would be balanced by shard across
all the machines not that ALL shards are balanced across the nodes,
regardless of index.
I'd love that as an option (to balance shards across the nodes by index),
as it's moving my two indexes to be split pretty much 1 index on 2
machines, 1 index on the other
2 machines. Since one of my indexes is a legacy index (we use it about
1/100th the other one), I'm still not utilizing my machines.
Now I'll just need to figure out how to trick it 
On Friday, March 16, 2012 7:10:05 PM UTC-7, John Cwikla wrote:
We have a 4 node cluster with 1 index, 25m docs or so. We have many
processes reading/writing to our cluster through
pyes which we've verified randomly chooses an initial node to talk to.Strange thing is, 3 of the 4 nodes are bored (0.25 load), 1 node is taking
the brunt -
40-50 load (16 core).We can't figure out why - any pointers? Anything I can do to debug what is
going on?