Our elasticsearch is in a red state. It was trying to allocate shards but some of our nodes were reporting many "too many open files" errors and the sharding process seemed to stall. I've seen that the recommended size for this /etc/init.d/elasticsearch variable is 65536 which is what we were set to. We increased it to 131070 and restarted the services. The sharding process restarted and is counting down but I have no way of knowing yet whether it will succeed.
What do you have your MAX_OPEN_FILES set to?