We are seeing some client side "retries exhausted" errors when we add a
node to our hbase cluster. Our settings are 4 tries at 1 second delay,
which seems to me like it should be enough time to avoid errors given
this post:
http://hadoop-hbase.blogspot.com/2014/08/hbase-client-response-times.html
When we decomission a node, we don't see errors. So can we assume the
errors occur because on node startup many regions move at once? Is there
a way to throttle this? We are running 0.94.20.
Thanks
Kireet
