> gossip traffic
That was imprecise, I should have said port 7000 instead of 'gossip'. I got
some help on irc, and it may have been a read repair, but not sure. What was
interesting was the internal load relative to anything happening externally
(120Mbps+ on 7000 compared to ~500Kbps on 9160)
A test 3 node cluster I'm running, managed by Priam, has 'suddenly' seen its
gossip traffic fly up, peaking at 120mbps. Looking at tcpdump it seems there is
simply a high rate of gossip traffic. There's nothing obviously wrong the
cluster itself; it's barely loaded, no markdowns/hints/gc and all
