Consul members fail - agent

Question

I am trying to do this tutorial--> tutorialConsul
When I run:
ikerlan#ikerlan-docker:~$ consul members
Node Address Status Type Build Protocol
ikerlan-docker 172.16.8.37:8301 alive client 0.3.0 2
172.16.8.191 10.0.2.15:8301 alive server 0.3.0 2
172.16.8.192 10.0.2.15:8301 alive server 0.3.0 2
172.16.8.193 10.0.2.15:8301 alive server 0.3.0 2
ikerlan#ikerlan-docker:~$ consul members
Node Address Status Type Build Protocol
ikerlan-docker 172.16.8.37:8301 alive client 0.3.0 2
172.16.8.191 10.0.2.15:8301 failed server 0.3.0 2
172.16.8.192 10.0.2.15:8301 alive server 0.3.0 2
172.16.8.193 10.0.2.15:8301 failed server 0.3.0 2
ikerlan#ikerlan-docker:~$ consul members
Node Address Status Type Build Protocol
ikerlan-docker 172.16.8.37:8301 alive client 0.3.0 2
172.16.8.191 10.0.2.15:8301 failed server 0.3.0 2
172.16.8.192 10.0.2.15:8301 alive server 0.3.0 2
172.16.8.193 10.0.2.15:8301 alive server 0.3.0 2
I can see the members alive, but if I run again some of that members failed, and then other ones alive and failed..
I think the leader, remove the members so they failed, and they try to connect but then they are removed.
I am working with virtualbox machines, and I see that consul takes all the machines with the same address, but I have configure consul to use eth1 interface that is the IP there appear in Node(name)
What could it be?
Thanks very much.

Answer 1

Two different nodes must not have the same address. This will confuse Consul, as Consul will use this address to connect to the nodes.
Use the bind_addr configuration option to ensure that consul uses the right interface and IP address on each machine consul doesn't pick the right address automatically. By default Consul picks the "first" private address it finds and this might not work in your setup.
The version you're running seems to be quite old (0.3.0). The current version is 0.6.0.