[Merged by Bors] - Packet filter cli option

[pawanjay176]

Issue Addressed

N/A

Proposed Changes

Adds a cli option to disable packet filter in lighthouse bootnode. This is useful in running local testnets as the bootnode bans requests from the same ip(localhost) if the packet filter is enabled.

[AgeManning]

hmm.... I dont recall banning localhost in the packet filter. Are you sure disabling the packet filter resolves this? I think the packet filter just limits requests and will ban an ip if it spams requests, which is probably what you're doing on smaller testnets.

Also, on smaller testnets you should set the --target-peer paramter to the expected number of nodes to prevent continuous discovery requests which will lead to discovery spamming

[pawanjay176]

It seems to be banning the last connecting localhost node from the logs

Aug 20 18:59:54.169 INFO Contact information                     multiaddrs: ["/ip4/127.0.0.1/udp/4242/p2p/16Uiu2HAm9youmPxCnW8WqdD9VNTqcTHfrd3PtDFsMEgsFZt7tkSj", "/ip4/127.0.0.1/tcp/4242/p2p/16Uiu2HAm9youmPxCnW8WqdD9VNTqcTHfrd3PtDFsMEgsFZt7tkSj"]
Aug 20 18:59:54.169 INFO Adding bootnode                         node_id: 0xaf62..e250, peer_id: 16Uiu2HAm9youmPxCnW8WqdD9VNTqcTHfrd3PtDFsMEgsFZt7tkSj, address: Some(127.0.0.1:4242)
Aug 20 18:59:54.169 INFO Discv5 Service started 
Aug 20 18:59:54.170 INFO Server metrics                          requests/s: 0.00, active_sessions: 0, connected_peers: 0
Aug 20 18:59:56.455 WARN Banning IP for excessive requests: 127.0.0.1 
Aug 20 19:00:04.170 INFO Server metrics                          requests/s: 2.00, active_sessions: 3, connected_peers: 3

This is on a local testnet with 4 beacon nodes all of them running with --target-peers 3.

[michaelsproul]

^I've observed the same as Pawan when running local testnets

[AgeManning]

Oh yeah. I think I added a limit per IP. So if all your nodes are under a single IP, its going to ban that IP.

Least the filters are working correctly.

I guess we should set this CLI flag for all the nodes on the testnet, not just the bootnodes, because the other nodes will also ban localhost from excessive traffic.

[pawanjay176]

I guess we should set this CLI flag for all the nodes on the testnet, not just the bootnodes, because the other nodes will also ban localhost from excessive traffic.

I think this should only happen after > 10 localhost beacon nodes which is sufficient for CI atleast (doppleganger tests)

lighthouse/beacon_node/eth2_libp2p/src/config.rs

Line 145 in 5b8436e

.filter_max_nodes_per_ip(Some(10))

Can add it in if required though :)

[AgeManning]

That config says that we allow up to 10 people to talk to us per IP.

However in the packet filter, we set default rate limits. There is a limit of how many messages we can receive per IP. See here: https://github.com/sigp/discv5/blob/master/src/config.rs#L106

The default value is on average 9 per second. This means per IP we only allow 9 requests per second before we ban the IP. So even though we allow 10 peers, on a small network, they will hit us with lots of requests such that we ban the IP from excessive requests.

So I think other nodes will also hit this limit if everyone is under the same IP.

[pawanjay176]

Ohh yeah that makes sense. Added the flag in beacon node as well in 2353618

[AgeManning]

Looks good to me

[AgeManning]

bors r+

[bors]

Pull request successfully merged into unstable.

Build succeeded:

• arbitrary-check
• cargo-audit
• cargo-fmt
• cargo-udeps
• check-benchmarks
• check-consensus
• clippy
• debug-tests-ubuntu
• dockerfile-ubuntu
• ef-tests-ubuntu
• eth1-simulator-ubuntu
• no-eth1-simulator-ubuntu
• release-tests-ubuntu
• release-tests-windows
• state-transition-vectors-ubuntu