#ls2 (irc2p) | #ls2 I2P Protocol Development Discussion - Next Meeting Monday, October 9th 6:30 UTC

orignal good idea

orignal will do the same

zzz here's where I sent a message to i2pd, retransmitted 3x, gave up after 10 seconds, then got acks for all of them 1 1/2 seconds later:

zzz 07-09 13:36:19.723 WARN [acket pusher] outer.transport.udp.PeerState2: Message expired: OB Message 2046472889 seq 84 type 23 size 2122 fragments: 2 volleys: 4 lifetime: 10044 unacked fragments: 0 sizes: 1417 705 send counts: 4 1 to: [2001:470:1f12:e56:0:0:0:2]:24768 BpATV4 OB2 recvAge: 10s sendAge: 10s sendAttemptAge: 3025ms sendACKAge: 10s lifetime: 117m RTT: 17 RTO: 8000 MTU: 1308 LMTU: 1500 cwin: 1500 acwin: 5137

zzz SST: 2744 FRTX? false consecFail: 1 msgs rcvd: 6 msgs sent: 85 pkts rcvd OK/Dup: 95/0 pkts sent OK/Dup: 91/4 IBM: 0 OBQ: 0 OBL: 0 theyRelayToUsAs: 3696531151

zzz 07-09 13:36:21.246 WARN [ handler 1/1] outer.transport.udp.PeerState2: Acked but not found in outbound messages: OB Message 2046472889 seq 84 type 23 size 2122 fragments: 2 volleys: 4 lifetime: 11567

zzz 07-09 13:36:21.246 WARN [ handler 1/1] outer.transport.udp.PeerState2: Dup ACK of fragment 0 of OB Message 2046472889 seq 84 type 23 size 2122 fragments: 2 volleys: 4 lifetime: 11567

orignal so what's your point?

orignal you think I miss packets?

zzz no. you got all the packets but never sent an ack

zzz until 11.5 seconds later

orignal what was after 11.5 secs?

orignal Just ack?

orignal 3 different acks?

zzz don't know, not logged

orignal or Ack with I2NP block?

orignal my understanding

orignal I have sent Acks but they didn't reach you

orignal after 11.5 I tried to send a new message

orignal with Ack block

zzz maybe

orignal then how can I fix it?

zzz just mentioning, in case you have a bug where you're not sending the acks.

zzz I will look for more delayed acks and see if there's a pattern

zzz you send an ack of the new packet even if the I2NP message inside is a duplicate, right???

orignal no

orignal I only send ack for a packet I didn't receive before

orignal and if a packet contains I2NP or fragments

zzz the distinction is between 'packet I didn't receive before' and 'packet containing i2np msg or fragment I didn't receive before'

zzz sounds like you do it right

zzz will keep looking

orignal let me change it

zzz so you don't do it right?

orignal myabe I should send Ack again if I receive duplicate

orignal need to check SSU1 code

zzz example:

zzz I send packet 1 with i2np message 12345

zzz you send ACK 1

zzz ACK 1 is lost

zzz I send packet 2 with i2np message 12345

zzz do you send ACK 2 or not?

orignal and I don't send ack

orignal wait

orignal if i's packet 2 I send ack

zzz ok good, you do it right

orignal but there is something else I don't do right

orignal I receive packet 1 and handle it

orignal I receive packet 2 and also handle it

orignal with the same msg 12345

R4SAS BpAT ... is mine

R4SAS rebuilt 2 days ago

zzz ok. is the internet reliable for it? It seems to have more errors than most routers

R4SAS ofcourse reliable, this is dedicated server

zzz ok, will keep an eye on it

zzz good to know

R4SAS ff + transit

R4SAS Routers: 5393 Floodfills: 1615 LeaseSets: 138

R4SAS Client Tunnels: 290 Transit Tunnels: 1952

zzz definitely seeing a lot of late acks from i2pd, after I've already timed out

zzz including 2RRY

zzz I have timestamps if you have the logs for it orignal

zzz R4SAS, do you have logs on BpAT for individual packets and acks?

orignal no

orignal but I will check

orignal late acks for how long?

orignal and 2RRY is not a good example

zzz as I said above, after my 10 second timeout

orignal as you know it really has netwrok issues

orignal anyway I will take a look what might cause it

R4SAS zzz: i have logs disabled, transit traffic generates too much messages\

R4SAS but i can enable them

zzz maybe a lower-bandwidth router would be better

zzz here's the router I got late acks from in last 12 hours: ImQC kyY2 2RRY YXEA BpAT = all i2pd

R4SAS I have ssu2-only router if you interested

orignal all R4SAS's ))

orignal but 2RRY

R4SAS нгз

R4SAS yup

R4SAS )))

zzz any of those low-bandwidth?

R4SAS none

R4SAS but I can disable transit

R4SAS lets do this on kyY2

zzz we need some router with low-level logging on

R4SAS disabling transit now

R4SAS I'll enable logging in 15 minutes

zzz maybe you introduced a bug recently? or maybe you two are running all the high-bandwidth routers? wondering why all the routers doing this are yours...

zzz also, we really need the termination issues fixed so we can reduce the number of packet failures

R4SAS most of them are high-bandwidth

R4SAS I'll rebuild with new commits today

orignal my that home ipv6 is low bndwidth

orignal zzz I have more routers

orignal that one with family=gostcoin and I rebuilt it yesteraday

orignal I'm working on termination issue now

zzz ok. missing termination and late acks are the two causes of message loss I'm seeing right now

zzz so many routers R4SAS, maybe you should set a family )))

orignal I will set gotscoin to 2RY

orignal always forget

R4SAS I'm not using family now because it will break ssu2 testing

zzz yeah, it's good for the high bandwidth ones, so we don't put them in the same tunnel

R4SAS maybe will add later

zzz let's focus on getting down to zero lost messages, or as close as we can get

R4SAS stopped kyY2, rebuilding

zzz make sure you're logging lost messages at a high level too so you can track it

orignal his routers have ygg address ending with 333 )))

R4SAS ))

zzz =zzz )

zzz restarting mine too

orignal old joke

orignal and remeber you will be prosecuted for your nick in some "democratic" countries now ))

zzz great

zzz >> teminate session after 5 unacked resends

zzz is that for one i2np message?

zzz or 5 different i2np messages?

orignal for packets

orignal I resend packets not messages

orignal and this 5 have exponential timeout

zzz that might be a litte drastic? we don't fail the session for one I2NP message, but for 5 consecutive I2NP messages that fail (each one retransmitted 4x)

zzz (same as in SSU 1 for us)

R4SAS started with debugging

zzz ours may not be drastic enough, 5 I2NP may be too high

orignal I will change

orignal also check size of unconrimed packets

orignal btw it might cauased delayed acks

orignal because I don't change Ack block when resend

orignal ofc it's 5 packets but "resendable"

orignal e.g. with I2NP message inside

orignal *5 resends