MQTT was restarted (up, but missing env vars for sphinx-auth, which mqtt uses for authorization of new clients)
a couple hours later the EC2 was dead
big spike in EC2 EBS reads.... possibly relay trying over and over to hit MQTT and getting 401... and for some reason loading the sqlite again and again? Or maybe crashing and restarting over and over?
We saw swarm4 relay crash when MQTT server reconnected (swarm4 is using
use-hd-keys
). Maybe there is a bug there?