Maximum Peers Reached, Connection To Private Node Dropped

Question

At 99 peers, my node 7a877820 (2019-12-11 15:02:03 +0100) starts getting

Dec 17 06:48:30 - p2p.connection-pool: Swap to x.x.x.x.:9732 failed: Error:
Dec 17 06:48:30 - p2p.connection-pool:           Too many connections.

But config does not have a max of 100 specified:

{ "rpc": { "listen-addrs": [ "127.0.0.1:8732" ] },
  "p2p":
    { "bootstrap-peers":
        [ "boot.tzbeta.net", "dubnodes.tzbeta.net:9732",
          "franodes.tzbeta.net:9732", "sinnodes.tzbeta.net:9732",
          "nrtnodes.tzbeta.net:9732", "pdxnodes.tzbeta.net:9732",
          "a.b.c.d:9732" ], "listen-addr": "0.0.0.0:9732",
      "limits":
        { "connection-timeout": 10, "min-connections": 15,
          "max_known_points": [ 240, 180 ],
          "max_known_peer_ids": [ 240, 180 ] } },
  "log": { "output": "/home/tezos/.tezos-node/tezos-node.log" },
  "shell":
    { "chain_validator": { "bootstrap_threshold": 2 },
      "history_mode": "archive" } }

I find 100 peers enough, but why does it consider too many peers at 100 if the config has no such limit? Is 100 the default max?

More importantly, it disconnects my private node, a.b.c.d:9732. The private node is set to run on cron tezos-admin-client trust peer id12345abcde, where id12345abcde is the id of public node with 100 connections. I don't think trust peer means stay connected to this peer, simply tells the private node to accept connections from node id id12345abcde, but the bottom line is: how do I ensure that the public node doesn't drop the connection to the private node when it reaches max connected nodes?

Answer 1

If you started your private node with eg.

[...] --net-addr 127.0.0.1:9876 [...]

add the following peer to the public node so that it trusts the private node:

[...] --peer 127.0.0.1:9876 [...]

The function that kills some connections when there are too many does not kill connections that are both in private mode and trusted.

See:

https://gitlab.com/tezos/tezos/blob/master/src/lib_p2p/p2p_maintenance.ml#L266 https://gitlab.com/tezos/tezos/blob/master/src/lib_p2p/p2p_maintenance.ml#L216

Answer 2

It seems that it is currently not possible to ensure the peer stays connected to your private node. See this and this.

So I created a mitigation script to trust and connect to the private node if it is not connected.