Cluster does not survive a vernemq restart of one server.

[lbthomsen]

From IRC discussion:

[18:01] Steps to provoke:
[18:02] 1. clean out /var/lib/vernemq completely on both systems with vernemq stopped
[18:02] 2. Start vernemq on both systems
[18:02] 3. One mosquitto_sub subscribed to a topic on first server
[18:02] 4. One mosquitto_sub susscribed to same topic on second server
[18:03] 5. Do the cluster join and verify both nodes report ok
[18:03] 6. Do a mosquitto_pub to one server - both subscribed servers receive
[18:03] 7. Do a mosquitto_pub to second server - both subscribed servers receive
[18:04] 8. Restart second server (vernemq restart)
[18:04] 9 Repeat 6-7
[18:04] Now it only works from server 2 to server 1
[18:04] Both servers still report ok on the cluster
[18:05] And it has not recovered in the time it took to write the above :)

[ioolkos]

Confirmed this. Looks like a bug in the re-connect logic from server1.