Thanks @ingenthr.
a. I can’t run cbcollect_info as it fails the servers, not sure why. They just lose connection and fail.
b. It seems that there was indeed problem with the memory although I’m not sure what. I’ve changed the servers to r3.xlarge (and by that had multiplied the available RAM) and the problem solved.
This problem started after a failed attempt to add nodes and rebalance. The nodes failed and since then the beam.smp and memcache processes started to consume a lot of memory.
c. We are running Ubuntu 14.04.
What is the beam.smp process?
Should we preserve an amount of memory for it? How much?
Thanks again.