listnetworkevents: keep a log of all connect/disconnect/failed-connect and pings #8558
Conversation
rustyrussell
force-pushed
the
guilt/networkevents
branch
from
6c0e9ed to
4eb19d5
Compare
|
Note: failed-connect should note if it's unreachable because there are no addresses using a protocol we support! |
rustyrussell
force-pushed
the
guilt/networkevents
branch
from
4eb19d5 to
b14d27e
Compare
rustyrussell
force-pushed
the
guilt/networkevents
branch
4 times, most recently
from
3039eb8 to
fe4ada6
Compare
There was a problem hiding this comment.
I think there's a problem with one of the sql insert statements. But otherwise this looks great. I think it will be really useful to automatically log all recent connections.
Do you think it would be worth timing out the ping and logging a lack of pong response, or would that be more problematic than it's worth? I'll try to test this out tomorrow.
connectd/connectd.c
Outdated
| * in progress right now) */ | ||
| if (connect->conn) | ||
| io_set_finish(connect->conn, NULL, NULL); | ||
| connect_reason = tal_steal(tmpctx, connect->reason); |
There was a problem hiding this comment.
So an inbound connection when we were also wanting to connect to the peer will still be attributed with our connection request reason?
There was a problem hiding this comment.
Agreed. If incoming, we should say "incoming connection while we were trying to connect" instead.
connectd/connectd.c
Outdated
| connect_time_nsec = time_to_nsec(timemono_since(connect->start)); | ||
| tal_free(connect); | ||
| } else { | ||
| connect_reason = ""; |
There was a problem hiding this comment.
Would it be helpful to explicitly label an "unrequested inbound connection" here?
…ults. Signed-off-by: Rusty Russell <[email protected]>
Signed-off-by: Rusty Russell <[email protected]>
Signed-off-by: Rusty Russell <[email protected]>
… *try* to connect to. This is important: if it's tor-only and we don't have a proxy, we will fail to connect, but it's no indication that the node is unreachable. Same with IPv6. Signed-off-by: Rusty Russell <[email protected]>
Changelog-Added: JSON-RPC: `wait` now has `networkevents` subsystem. Signed-off-by: Rusty Russell <[email protected]>
Signed-off-by: Rusty Russell <[email protected]>
Signed-off-by: Rusty Russell <[email protected]>
Signed-off-by: Rusty Russell <[email protected]>
Signed-off-by: Rusty Russell <[email protected]> Changelog-Added: JSON-RPC: `sql` now supports the `networkevents` table.
Changelog-Added: JSON-RPC: `delnetworkevent` to delete from listnetworkevents. Signed-off-by: Rusty Russell <[email protected]>
We also document this in the listnetworkevents command itself. The test_autoclean_once was getting repetitive, so I cleaned that up too. Signed-off-by: Rusty Russell <[email protected]> Changelog-Added: `autoclean` will remove networkevents after 30 days by default.
rustyrussell
force-pushed
the
guilt/networkevents
branch
from
fe4ada6 to
ac80f2f
Compare
2 participants
Combined with gossipd connecting to random peers, this gives us an indication of peer latency. This might be useful to someone.