Emit event for User's NAT Type i.e. Hard NAT or Easy NAT #1042

aarshkshah1992 · 2021-01-27T12:29:10Z

We should emit an event informing the user of the user's NAT type for better user experience for hole punching/testing Hole punching in the wild.

TODO

Merge Event for user's NAT Device Type: Tell user if the node is behind an Easy or Hard NAT go-libp2p-core#173 AND Update deps.

vyzo

I am a little sceptical about the activated address heuristic; if we have one, then yes we are obviously behind an easy NAT.
But if we don't we might still be behind a penetrable NAT, especially with UDP.

aarshkshah1992 · 2021-01-27T12:52:42Z

@vyzo

We don't determine the NAT to be hard simply if we don't have an activated address. We determine the NAT to be hard when four different peers report four different observed addresses on outbound connections. I think this should be good enough to conclude that the NAT is bad, no ?

We can augment the code to report NAT types separately for UDP & non-UDP connections if that's what you mean.

vyzo · 2021-01-27T13:30:19Z

Yeah, we should probably do that.

p2p/protocol/identify/obsaddr.go

willscott · 2021-01-28T17:21:51Z

p2p/protocol/identify/obsaddr.go

+	oas.emitUDPTypeUnlocked(allObserved)
+}
+
+func (oas *ObservedAddrManager) emitTCPNATTypeUnlocked(addrs []*observedAddr) {


this and the subsequent emitUDPNATTypeUnlocked are doing basically the same thing. can you factor out a common method that takes an additional parameter (ma.P_TCP vs ma.P_UDP)?

This is done.

aarshkshah1992 · 2021-02-01T11:57:00Z

@willscott I've addressed both your comments. Have also tested this on my machine for both TCP & QUIC and it looks good.
Thanks !

willscott

one minor nit, and you need to get the dependencies merged and go.mod pulled up to versions before merging, but this seems ready.

willscott · 2021-02-01T15:51:45Z

p2p/protocol/identify/obsaddr.go

+//
+// Please see the documentation on the enumerations for `network.NATDeviceType` for more details about these NAT Device types
+// and how they relate to NAT traversal via Hole Punching.
+func (oas *ObservedAddrManager) emitNATTypeUnlocked() {


why is this method named Unlocked? maybe this is observeAllNATTypes and the next one is observeSpecificNATType? the naming of the two feels not quite representative of what they do.

yeah what's with the name? Is there a locked variant?

Have fixed this.

aarshkshah1992 · 2021-02-15T12:05:15Z

cc @vyzo

vyzo · 2021-02-17T17:47:48Z

p2p/protocol/identify/obsaddr.go

+//
+// Please see the documentation on the enumerations for `network.NATDeviceType` for more details about these NAT Device types
+// and how they relate to NAT traversal via Hole Punching.
+func (oas *ObservedAddrManager) emitNATTypeUnlocked() {


yeah what's with the name? Is there a locked variant?

vyzo · 2021-02-17T17:49:39Z

p2p/protocol/identify/obsaddr.go

+
+// returns true along with the new NAT device type if the NAT device type for the given protocol has changed.
+// returns false otherwise.
+func (oas *ObservedAddrManager) emitNATType(addrs []*observedAddr, protoCode int, transportProto event.NATTransportProtocol,


This could range from a little to very expensive.
How often do we call it?
Can we skip it altogether if we are not NATted?

@vyzo

What makes this expensive ?

We call it every-time we have a new observed address but this function will return as soon as it sees any activated address.

The problem with skipping it altogether if we are not NATT'd is that AutoNAT dosen't take into account TCP vs UDP whereas we want to emit separate events for both.

Do you mean to say that we should emit this event for both TCP & UDP ONLY if reachability is private ?

Imagine a big public peer with thousands of connections, this could become problematic.

And yes, this event is only relevant if we have private reachability and can/should be skipped altogether if we are public.

aarshkshah1992 · 2021-02-18T11:33:34Z

@vyzo

Addressed your review.
Rebased & updated deps.

Please can you take a look ?

vyzo

One bug and a bunch of stylistic issues...

vyzo · 2021-02-18T17:09:51Z

p2p/protocol/identify/obsaddr.go

+		case evt, ok := <-subChan:
+			if !ok {
+				subChan = nil
+			}
+			ev := evt.(event.EvtLocalReachabilityChanged)
+			oas.reachability = ev.Reachability


this will panic when the channel gets closed, as evt will be nil.

oops :/ Fixed.

vyzo · 2021-02-18T17:11:50Z

p2p/protocol/identify/obsaddr.go

+	for k := range oas.addrs {
+		allObserved = append(allObserved, oas.addrs[k]...)
+	}


Index again when you mean to iterate.... It's more idiomatic to write this as

for _, addrs := range oas.addrs { allObserved = append(allObserved, addrs...) }

More pertinent however, do we really need this copy?

well, it's probably a micro-optimization to fix the copy. Have fixed the looping .

no you haven't, it's still there.

vyzo · 2021-02-18T17:13:38Z

p2p/protocol/identify/obsaddr.go

+	for i := range addrs {
+		oa := addrs[i]


and here again:

for _, oa := range addrs { ... }

vyzo · 2021-02-18T17:14:40Z

p2p/protocol/identify/obsaddr.go

+		// An observed address on an outbound connection that has ONLY been seen by one peer
+		if now.Sub(oa.lastSeen) <= oas.ttl && oa.numInbound == 0 && len(oa.seenBy) == 1 {
+			cnt++
+			for s, _ := range oa.seenBy {


you don't need the , _.

aarshkshah1992

@vyzo

Have made the changes.

vyzo

needs a go mod tidy; also, fix dat loop!

vyzo · 2021-02-19T08:41:55Z

go.mod

+	github.com/google/gopacket v1.1.18 // indirect
+	github.com/google/uuid v1.1.2 // indirect


what's that crap popping up? Do a go mod tidy.

vyzo · 2021-02-19T08:43:49Z

p2p/protocol/identify/obsaddr.go

+	for k := range oas.addrs {
+		allObserved = append(allObserved, oas.addrs[k]...)
+	}


no you haven't, it's still there.

aarshkshah1992 · 2021-02-19T09:28:55Z

@vyzo This is ready.

vyzo

alright, let's merge this.

aarshkshah1992 requested review from vyzo, Stebalien and willscott January 27, 2021 12:29

aarshkshah1992 mentioned this pull request Jan 27, 2021

Project Flare(decentralised Hole Punching) Phase1 Meta Issue #1039

Closed

23 tasks

vyzo reviewed Jan 27, 2021

View reviewed changes

aschmahmann mentioned this pull request Jan 28, 2021

feat: basic ipfs network diagnostics ipfs/kubo#7068

Draft

aarshkshah1992 mentioned this pull request Jan 28, 2021

Update IPFS network diagnostics with NAT type #1043

Closed

willscott reviewed Jan 28, 2021

View reviewed changes

willscott approved these changes Feb 1, 2021

View reviewed changes

vyzo reviewed Feb 17, 2021

View reviewed changes

aarshkshah1992 added 12 commits February 18, 2021 16:52

emit NAT device type

10d4231

unit tests

b2ba697

emit event for NAT Device Type

47bbd95

updated deps

2bffd08

refactor tcp-udp NAT events

80e622b

go mod tidy

15af54c

update deps

7e227b8

fix multiple events

8085cda

fixed naming

84345fa

emit event ONLY if reachability is private

ef82c43

address review

8f6554b

fix test

df2b0be

aarshkshah1992 force-pushed the feat/nat-type branch from 5bcd646 to df2b0be Compare February 18, 2021 11:28

vyzo reviewed Feb 18, 2021

View reviewed changes

changes as per review

d722ec3

aarshkshah1992 commented Feb 19, 2021

View reviewed changes

vyzo reviewed Feb 19, 2021

View reviewed changes

final review

8f91710

vyzo approved these changes Feb 19, 2021

View reviewed changes

aarshkshah1992 merged commit 69916ed into master Feb 19, 2021

aarshkshah1992 deleted the feat/nat-type branch February 19, 2021 09:44

Stebalien mentioned this pull request May 11, 2021

Release v0.14.0 #1090

Closed

27 tasks

aschmahmann mentioned this pull request May 14, 2021

Release v0.9.0 ipfs/kubo#8058

Closed

71 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Emit event for User's NAT Type i.e. Hard NAT or Easy NAT #1042

Emit event for User's NAT Type i.e. Hard NAT or Easy NAT #1042

aarshkshah1992 commented Jan 27, 2021 •

edited

Loading

vyzo left a comment

aarshkshah1992 commented Jan 27, 2021 •

edited

Loading

vyzo commented Jan 27, 2021

willscott Jan 28, 2021

aarshkshah1992 Feb 1, 2021

aarshkshah1992 commented Feb 1, 2021 •

edited

Loading

willscott left a comment

willscott Feb 1, 2021

vyzo Feb 17, 2021

aarshkshah1992 Feb 18, 2021

aarshkshah1992 commented Feb 15, 2021

vyzo Feb 17, 2021

vyzo Feb 17, 2021

aarshkshah1992 Feb 18, 2021

vyzo Feb 18, 2021

aarshkshah1992 Feb 18, 2021

aarshkshah1992 commented Feb 18, 2021

vyzo left a comment

vyzo Feb 18, 2021

aarshkshah1992 Feb 19, 2021

vyzo Feb 18, 2021

vyzo Feb 18, 2021

aarshkshah1992 Feb 19, 2021

vyzo Feb 19, 2021

vyzo Feb 18, 2021

aarshkshah1992 Feb 19, 2021

vyzo Feb 18, 2021

aarshkshah1992 Feb 19, 2021

aarshkshah1992 left a comment

vyzo left a comment

vyzo Feb 19, 2021

vyzo Feb 19, 2021

aarshkshah1992 commented Feb 19, 2021

vyzo left a comment

		github.com/google/gopacket v1.1.18 // indirect
		github.com/google/uuid v1.1.2 // indirect

Emit event for User's NAT Type i.e. Hard NAT or Easy NAT #1042

Emit event for User's NAT Type i.e. Hard NAT or Easy NAT #1042

Conversation

aarshkshah1992 commented Jan 27, 2021 • edited Loading

TODO

vyzo left a comment

Choose a reason for hiding this comment

aarshkshah1992 commented Jan 27, 2021 • edited Loading

vyzo commented Jan 27, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aarshkshah1992 commented Feb 1, 2021 • edited Loading

willscott left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aarshkshah1992 commented Feb 15, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aarshkshah1992 commented Feb 18, 2021

vyzo left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aarshkshah1992 left a comment

Choose a reason for hiding this comment

vyzo left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aarshkshah1992 commented Feb 19, 2021

vyzo left a comment

Choose a reason for hiding this comment

aarshkshah1992 commented Jan 27, 2021 •

edited

Loading

aarshkshah1992 commented Jan 27, 2021 •

edited

Loading

aarshkshah1992 commented Feb 1, 2021 •

edited

Loading