Dropped off after few seconds on every server, 255 ack problem

oberonIT · October 2015

Hello,
(I tried all the usual stuffs, playing without mods, deleting %appdata%/natural selection 2/, deleting the userdata, reinstalling, integrity check and stuffs, and still get this)
When I connect to a server I get to login normally for a few seconds, even join teams, spawn and listen to comms, then I get the "disconnected"/unplugged icon on the left, and all the console says is a bunch (around 50) of:
[ 81.759] MainThread : 37.59.11.151:27015 [Client]: Ack 255 outside sequence window 45 - 173
Until I eventually time out.
The port and the ack windows vary slightly, but at the end it's the same on every server. I can ping and see the servers, I tried no-ns2+ servers, I changed dns.
I have the tech_support.zip but I'd rather not drop it on the forum.

(another example: [ 91.819] MainThread : 144.76.119.131:14015 [Client]: Ack 255 outside sequence window 46 - 174)

DC_Darkling · October 2015

try it on the taw pub and im willing to compare server side logs.
Need a time & date obviously.

oberonIT · October 2015

Just tried. So the first 2 times I got dropped, the third I didnt. And then the 4th I got dropped again. Just turned off antivirus and windows firewall btw.

oberonIT · October 2015

(time and date of the post so you get it in your timezone)
nevermind the hour get eaten, it was about 23pm (gmt+1)

IronHorse · October 2015

Are you running ANYTHING else on port 27015?
Be very sure, because people who are running TF2 servers on the same connection have had the same issue.

DC_Darkling · October 2015

You do realise TAW has 8 running servers yes? :P

Ok as your forum name does not match any of the ingame name of folk who were on the taw server, im just guessing the following. I need more info like your game name, ip or ID to make sure it was you.
[4636.167] Main : ClientIPHere : Ack 203 outside sequence window 255 - 127

That line is shown in the server log in oceans of it. (obviously with the ip in it)
So far I see the person in question could finally connect.

oberonIT · October 2015

Ah it did match but then I changed suspecting it was a hive related problem, it was Neur0mante at the time of the tests. I swear I could only see the UK one in the server list ^_^ it's not proxed and the port is even opened, did the IP included a port? Anyway I have the same problem no matter what port the server is answering through

DC_Darkling · October 2015

Yes the player I am talking about was named Neur0mante.

Your client port on the Ack spam was 57676. (Ack 203 outside sequence window 255 - 127)
Your client port upon connect was 60658.
You then had a Ack spam of port 63694. (Ack 212 outside sequence window 255 - 127)
The server then threw a error in-between of which im unsure if its related. ([4877.968] Main : Error: Exceeded maximum number of snapshots)
Then you disconnected by normal quit.

You then connected on port 63694.
You did a suicide in-game.
You quit.

Connect on 60886.
Ack spam on 62498. (Ack 228 outside sequence window 255 - 127)
Snapshot error ([5379.989] Main : Error: Exceeded maximum number of snapshots)
Disconnect on timeout.

Hmm, you make me wonder about the snapshot error.. haven't seen that one yet but that does not mean it is related.
ps.. its perfectly normal to have variable client ports.

IronHorse · October 2015

Can you attempt connecting to his servers after typing net_log 1024 in the console?
Then provide us with your techsupport.zip file that's generated by running techsupport.exe from your NS2 folder. (ensure ns2 is closed when you run it)

DC_Darkling · October 2015

Want me to do that server side also @ironhorse ?
Will that help?
I dont wanna do it for days obviously if it fills logs that much.

IronHorse · October 2015

Not needed, but thanks

oberonIT · October 2015

So I was able to connect to the taw[uk] server sometimes, like 1/5 or so, the others not so much. When I'm able to connect there's no trace of the ack problem. The log.old is pretty clear, in the second one I mistakenly typed net_log 1014 which gave a bunch of different packet informations, but you may wanna check those too.

DC_Darkling · October 2015

Let me know if I need to compare server side logs @ironhorse.
Usually got till sunday before we clean those up at maintenance.

IronHorse · October 2015

[414.929] MainThread : Write index 9, lastRec 1, ppt 0, loss 89 @ time 414.929
[400.794] MainThread : Packet serial 1 out of order, last 1, diff 0, outOfOrder 1, outLoss 120, inLoss 89

You seem to be encountering a fair amount of lost packets. Ending in a timeout.

First step:
Use http://tools.pingdom.com/ping/ to test the packet loss to the server IP that you are having difficulty with (exclude the port number, and use the resolved address, so for instance phenomundertaker.ddns.net resolves to 31.51.47.102)
Then post the results / picture here for us to look at. This will test your connection to the server without NS2 being involved.
Please run it at least 8 times in a row, and only when you are having difficulty connecting. Packet loss can be a result of many factors, but mostly congestion, so time of day etc does matter.

If you see packet loss there, please look for any loss between your router and your computer. Then post your results here as well.

Also try running pingtest.net just for kicks.

Thanks

EDIT: attempting to ping that server results in a timeout for myself as well, even though I can join in game just fine. Hmmm investigating...

EDIT2 : Looks like server is blocking ICMP - that's why it failed to be pinged. But the good news is that the CDT already knows what is causing this for you.
It has to do with bad routing and the server starts assuming the client got disconnected too early, then the server treats the next packets as new connection while the client treats them as the old session.

So expect a fix soon, and thanks for the data.

BabblerKing · October 2015

What? I had sent that stuff in as early as pre 276.1
good work

oberonIT · October 2015

That was awfully quick, thanks for the help. I'm removing the dxdiag just for tidiness, glad to contribute smashing bugs

matso · October 2015

Yea, DC_Darkling supplied me with the log. The crucial part here is these two lines:

Client connecting (80.181.87.223:61984)
[17632.463] Main : 80.181.87.223:64604 : Ack 225 outside sequence window 255 - 127

This shows that the client changes its port during the connection, which is a no-no - the server uses the (host,port) combo to identify a client. So to the server, the old client stops talking and a new client connects that behaves really badly.

The most likely explanation for this is that you are connecting to the internet through a NAT gateway; it translates the local internet address you use to a public address usable on the actual internet.

So the host ip is the NAT gateway ip address, not yours, and the port is a temporary port setup by the NAT gateway to receive replies to the UDP packet you sent.

What seems to happen is that the NAT gateway after a minute or so decides to drop the binding between your UDP socket to the NAT gateway socket, and then when the next packet arrives from you, it gets a new binding; ie a new port. Which breaks things ... badly.

How you connect to the internet would be interesting to know.

I think the network protocol can be updated to handle it though; just add an id to each packet that identifies the peer and remap the return address when packets starts coming in from a new source.

Doable.

oberonIT · November 2015

The router is a vdsl telecom(italy) branded router, the available settings are really barebone. Since it's also decoding the telephone line I cant really ditch it.
http://www.hwupgrade.it/forum/showthread.php?t=2587383 this guy. Can't find an english page for it.
A friend of mine has the same router and no problem tho, so it could be on my end.

Dropped off after few seconds on every server, 255 ack problem

Comments