You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently, the network has more provers than clients / validators. As provers don't sync with the network, it makes bootstraping and syncing hard for nodes:
Most of the peers of bootstrap nodes are provers, and as bootstrap nodes currently are the only canonical block source, sometimes new blocks can't promptly propagate to other network nodes [1];
If client nodes has most of their peers as provers, it will limit their view into the network and can't have many peers to sync from.
Also, as the bootstrap nodes has the same max peers as normal clients, a new node will have difficulties connecting to any one of them to do the "bootstrap" (discovering nodes). Moreover, it seems some peers of the bootstrap nodes are hardcoded internal nodes (presumably the validators run by the Aleo team), the actual max peers for them are smaller than normal client nodes.
The proposal is:
Don't count provers when counting the connected peers, or count them separately so nodes could connect to more clients / validators - provers don't sync so their connections should use less resources than other peers;
Increase the max peers of bootstrap nodes for now to allow more nodes to connect to them.
[1] This is also partly because we only rely on polling to sync blocks instead of broadcasting new blocks, but there are still syncing issues if a node is behind the network and need to catch up
The text was updated successfully, but these errors were encountered:
馃挜 Proposal
Currently, the network has more provers than clients / validators. As provers don't sync with the network, it makes bootstraping and syncing hard for nodes:
Also, as the bootstrap nodes has the same max peers as normal clients, a new node will have difficulties connecting to any one of them to do the "bootstrap" (discovering nodes). Moreover, it seems some peers of the bootstrap nodes are hardcoded internal nodes (presumably the validators run by the Aleo team), the actual max peers for them are smaller than normal client nodes.
The proposal is:
[1] This is also partly because we only rely on polling to sync blocks instead of broadcasting new blocks, but there are still syncing issues if a node is behind the network and need to catch up
The text was updated successfully, but these errors were encountered: