Incompatible peers should be marked bad #3081
Description
ABCI app (name for built-in, URL for self-written if it's publicly available):
gaia v.0.29.1
What happened:
Log entry on freshly spun up genki-4002 node:
Jan 04 07:33:28 i-0dbf25d1b1adc068c gaiad[15884]: E[4016-01-04|07:33:28.439] Dialing failed module=pex addr=1e825baf672f3193f7457ff4ce9c3865dc21057c@51.15.121.63:26656 err="incompatible: Peer is on a different network. Got game_of_stakes_3, expected genki-4002" attempts=0
Jan 04 07:33:56 i-0dbf25d1b1adc068c gaiad[15884]: E[4016-01-04|07:33:56.319] Dialing failed module=pex addr=1e825baf672f3193f7457ff4ce9c3865dc21057c@51.15.121.63:26656 err="incompatible: Peer is on a different network. Got game_of_stakes_3, expected genki-4002" attempts=1
Jan 04 07:34:24 i-0dbf25d1b1adc068c gaiad[15884]: E[4016-01-04|07:34:24.931] Dialing failed module=pex addr=1e825baf672f3193f7457ff4ce9c3865dc21057c@51.15.121.63:26656 err="duplicate CONN<51.15.121.63:26656>" attempts=2
Jan 04 07:34:57 i-0dbf25d1b1adc068c gaiad[15884]: E[4016-01-04|07:34:57.927] Dialing failed module=pex addr=1e825baf672f3193f7457ff4ce9c3865dc21057c@51.15.121.63:26656 err="auth failure: secrect conn failed: read tcp 10.42.42.185:59088->51.15.121.63:26656: i/o timeout" attempts=3
Jan 04 07:35:27 i-0dbf25d1b1adc068c gaiad[15884]: E[4016-01-04|07:35:27.930] Dialing failed module=pex addr=1e825baf672f3193f7457ff4ce9c3865dc21057c@51.15.121.63:26656 err="auth failure: secrect conn failed: read tcp 10.42.42.185:59472->51.15.121.63:26656: i/o timeout" attempts=4
Jan 04 07:36:24 i-0dbf25d1b1adc068c gaiad[15884]: E[4016-01-04|07:36:24.932] Dialing failed module=pex addr=1e825baf672f3193f7457ff4ce9c3865dc21057c@51.15.121.63:26656 err="duplicate CONN<51.15.121.63:26656>" attempts=5
Jan 04 07:37:54 i-0dbf25d1b1adc068c gaiad[15884]: E[4016-01-04|07:37:54.960] Dialing failed module=pex addr=1e825baf672f3193f7457ff4ce9c3865dc21057c@51.15.121.63:26656 err="incompatible: Peer is on a different network. Got game_of_stakes_3, expected genki-4002" attempts=6
Jan 04 07:40:24 i-0dbf25d1b1adc068c gaiad[15884]: E[4016-01-04|07:40:24.929] Dialing failed module=pex addr=1e825baf672f3193f7457ff4ce9c3865dc21057c@51.15.121.63:26656 err="duplicate CONN<51.15.121.63:26656>" attempts=7
Jan 04 07:44:54 i-0dbf25d1b1adc068c gaiad[15884]: E[4016-01-04|07:44:54.959] Dialing failed module=pex addr=1e825baf672f3193f7457ff4ce9c3865dc21057c@51.15.121.63:26656 err="incompatible: Peer is on a different network. Got game_of_stakes_3, expected genki-4002" attempts=8
Jan 04 07:53:54 i-0dbf25d1b1adc068c gaiad[15884]: E[4016-01-04|07:53:54.961] Dialing failed module=pex addr=1e825baf672f3193f7457ff4ce9c3865dc21057c@51.15.121.63:26656 err="incompatible: Peer is on a different network. Got game_of_stakes_3, expected genki-4002" attempts=9
Relevant part of addrbook.json
{
"addr": {
"id": "1e825baf672f3193f7457ff4ce9c3865dc21057c",
"ip": "51.15.121.63",
"port": 26656
},
"src": {
"id": "ed6b6d5019563b40e81ae29c80c712fce7ae68f0",
"ip": "5.83.160.83",
"port": 26656
},
"attempts": 24,
"last_attempt": "2019-01-04T07:44:54.96000829Z",
"last_success": "0001-01-01T00:00:00Z",
"bucket_type": 1,
"buckets": [
18,
125,
130,
199
]
},
What you expected to happen:
Incompatible peers should be marked bad and "never" attempted again.
Couple if things to note:
- MarkBad is currently only used for failed connection attempts. Incompatible peers should also be marked bad.
- MarkBad in AddrBook simply removes a peer today, so it will be retried immediately once added again (via RPC/PEX).
- Bad peers should ideally be tracked using a flag and/or expiry date ("marked bad until..."), after which it can be removed from AddrBook. It can then be retried after a sensible time, if added again.
It might also be worth considering not including bad peers in PEX. But there are pros/cons.