Node dead after start [ DELTA 6 ] about casper-node HOT 2 CLOSED

Thanks for this report too :)

It seems that this is an effect of increasing the gossip_interval in the config to 300_000 (i.e. 5 minutes). We overlooked the fact that the node's consensus component requires there to have been some consensus messages within 5 minutes of the component being constructed or else it deems itself non-functioning and kills the node.

To receive consensus messages, the node must have some incoming connections (the peers listed via the /status endpoint could be incoming or outgoing ones). To get an incoming connection, the node must gossip its public address, and remote peers receiving this will then try and connect to it.

By only having peers gossip their addresses on a 5 minute interval, there's no chance for a node joining the network to establish incoming connections in time to satisfy the consensus requirement.

The simplest fix (which worked when we tried on one of our own nodes) is to reduce the config setting back down. We opted to change it to:

gossip_interval = 120_000

Manually making that config change and restarting the node should allow it to join successfully.

I'll leave this issue open until we fix that setting in our release packages.

from casper-node.

Fixed by #707

from casper-node.