feat: add control over autonat #176
Conversation
| ); | ||
| autonat.probe_address(info.observed_addr.clone()); | ||
| // Only probe one address at a time. | ||
| if self.active_address_probe.is_none() { |
There was a problem hiding this comment.
if there's already a probe running, then it looks like we drop this addr on the floor. Should we insert it into a queue to probe later?
There was a problem hiding this comment.
I thought about it but I don't think its needed. Everytime we connect to a peer it will report the observed address and trigger this code. So this logic is run very frequently. Additionally its typical to have only a single external address. In which case its likely that the in-progress address probe is the correct one. In cases where there are lots of different observed addresses its likely that NAT hasn't been setup correctly for the peer and so doesn't have an external address.
Maybe I could add this as a comment on in the code?
There was a problem hiding this comment.
Yeah, I think a comment clarifying this would be great
p2p/src/node.rs
Outdated
| // This logic is run very frequently because any new peer connection | ||
| // for a new observed address triggers this path. Its typical to have | ||
| // only a few external addresses, in which cases its likely that the | ||
| // in-progress address probe is one that will succeeded. |
There was a problem hiding this comment.
| // in-progress address probe is one that will succeeded. | |
| // in-progress address probe is one that will succeed. |
Prior to this change autonat would always be enabled. Now it can be disabled in cases where NAT is not being used. Additionally previously an unbounded number of probes would be running concurrently for all observed addresses. When an dynamic outbound NAT is used this results in many probes. Now we limit to one active probe at a time and also track failed external address so we do not repeatedly attempt to probe them.
nathanielc
force-pushed
the
fix/autonat-probe
branch
from
80fb3f8
to
648c07f
Compare
Prior to this change autonat would always be enabled. Now it can be disabled in cases where NAT is not being used.
Additionally previously an unbounded number of probes would be running concurrently for all observed addresses. When an dynamic outbound NAT is used this results in many probes. Now we limit to one active probe at a time and also track failed external address so we do not repeatedly attempt to probe them.