I’ve been experimenting with using dat to sync data from one machine to a cluster of machines. If it matters, the cluster is on a lan and the seed machine is connected to that lan over a VPN. Occasionally, the seed machine will go down for a few hours. During this time, the cluster successfully maintains the health of the swarm so long as one of the machines is always up. The problem I have is that when this seed machine comes back online, it oftentimes is never discovered by the cluster. If I kill one of the cluster machines to trigger a fresh sync in the cluster, this occasionally notices the seed machine and gets it connected to the cluster. Lately however, that hasn’t been working either. The only thing that seems reliable is killing all of the cluster dat nodes enforcing a re download of the dat archive. Because all of the cluster nodes are starting from scratch, they are forced to connect to the seed machine.
Is there some other way I can reliably encourage the cluster to reconnect to the seed machine without nuking the entire thing? Perhaps there is some way to hardcode an ip address or dat node?
It seems unlikely that this is a networking problem since the nodes have no trouble establishing connections with each other when starting from a fresh dat sync.