To run privately an ethereum cluster, the nodes have to find each other through a bootnode. The address of the bootnode has to be an IP address. When we run a bridged network we can set manually the ip address, but this option is not currently supported by the overlay network needed by swarm.
Therefore in our init script we run a simple resolution command based on the image name "bootnode". Problem was that the nodes were still unable to connect and no mining would happen.
After investigating, I've noticed that the address resolved with the name of the service was not consistent. Inside the docker container:
The right address was .4 where the system was answering , with .3 other containers were not able to connect. Therefore I change the bash script to fetch the bootnode ip by prepending the hostname with "tasks." like doing "host tasks.bootnode" instead of "host.bootnode" (you could use whatever command you like):
BOOTNODEIPADDRESS=$(host tasks.$3 | head -1|cut -d' ' -f4)
Therefore in our init script we run a simple resolution command based on the image name "bootnode". Problem was that the nodes were still unable to connect and no mining would happen.
After investigating, I've noticed that the address resolved with the name of the service was not consistent. Inside the docker container:
root@84cba6d65c3d:/geth# host ethnode4
ethnode4 has address 10.0.0.3
root@84cba6d65c3d:/geth# host tasks.ethnode4
tasks.ethnode4 has address 10.0.0.4
BOOTNODEIPADDRESS=$(host tasks.$3 | head -1|cut -d' ' -f4)
After this, the ethereum cluster started correctly
Comments