Problem with more than 2 LAN's connected...

Gamer's Internet Tunnel, formerly Gamer's IPX Tunnel

Problem with more than 2 LAN's connected...

Postby Viatrak » Thu Dec 30, 2004 6:31 pm

LAN 1 CONFIG
IP 69.141.228.XXX PORT 213 METHOD UDP notes: LAN 1 TO LAN 2
IP 24.98.232.XXX PORT 214 METHOD UDP notes: LAN 1 TO LAN 3
FORWARD 213-214

LAN 2 CONFIG
IP 68.162.40.XXX PORT 213 METHOD UDP notes: LAN 2 TO LAN 1
IP 24.98.232.XXX PORT 215 METHOD UDP notes: LAN 2 TO LAN 3
FORWARD 213-215

LAN 3 CONFIG
IP 68.162.40.XXX PORT 214 METHOD UDP notes: LAN 3 TO LAN 1
IP 69.141.228.XXX PORT 215 METHOD UDP notes: LAN 3 TO LAN 2
FORWARD 214-215

Ethernet II
Forward ARP
TCP
UDP
ICMP
ZLib Compression (tried with and without)
Also Match Source Port
Don't send routable (tried with and without)

When I connect all 3 lan's together I get HUGE packet spam (like 1k+ packets a second). To me it seems like the packets just keep going in a circle and never "die." No IP Conflicts, all firmware is upgraded, all same subnet, and any 2 lan's can play together fine.

Maybe I'm setting up the routing wrong, any ideas?

Thanks
Viatrak
 
Posts: 5
Joined: Thu Dec 30, 2004 6:23 pm

Postby Ark » Thu Dec 30, 2004 6:41 pm

What version of GIT are you using? Try 0.99 BETA 4 and make sure not to check the 'use old reforward prevention method' type option.
Ark
Site Admin
 
Posts: 2108
Joined: Sat Sep 13, 2003 4:21 pm

Postby Viatrak » Thu Dec 30, 2004 6:45 pm

I'm using beta 4, and the option is not checked... although I also did try the option.
Viatrak
 
Posts: 5
Joined: Thu Dec 30, 2004 6:23 pm

Postby Ark » Thu Dec 30, 2004 6:50 pm

What you have for the two UDP connections is correct, in fact, its probably exactly like the generic TCP/UDP wizard would configure it for.

Check in the Connection Status window and check the box "view known remote host list" to see a list of hosts GIT has received packets from. In the middle of the "packet storm" you are getting, copy this list of remote hosts from all 3 GITs and paste it here to compare them.
Ark
Site Admin
 
Posts: 2108
Joined: Sat Sep 13, 2003 4:21 pm

Postby Viatrak » Thu Dec 30, 2004 7:17 pm

LAN 1
00:0E:A6:46:19:6A - 25
00:0F:66:B7:06:5C - 25
00:11:2F:3F:29:AC - 25
00:02:E3:23:D8:DF - 25
00:0F:66:41:7B:D8 - 22
00:0C:41:A3:20:F6 - 9
00:0C:41:5F:0D:CA - 10

LAN 2
00:0C:41:A3:20:F6 - 27
00:04:23:59:90:EA - 30
00:02:E3:23:D8:DF - 30
00:0E:A6:6B:DD:C3 - 30
00:0E:A6:46:19:6A - 30
00:0D:88:2F:A8:B7 - 28
00:0C:41:5F:0D:CA - 14

LAN 3
00:0F:66:B7:06:5C - 9
00:0E:A6:6B:DD:C3 - 30
00:0D:88:2F:A8:B7 - 8
00:04:23:59:90:EA - 30
00:0F:66:41:7B:D8 - 13
00:11:2F:3F:29:AC - 30
Viatrak
 
Posts: 5
Joined: Thu Dec 30, 2004 6:23 pm

Postby Viatrak » Thu Dec 30, 2004 7:33 pm

The mac's of the people trying to the play the games are:

Network A MAC's
00:0E:A6:6B:DD:C3
00:04:23:59:90:EA

Network B MAC's
00:11:2F:3F:29:AC

Network C MAC's
00:0E:A6:46:19:6A
00:02:E3:23:D8:DF

The PC's running GIT are:

Network A (GIT PC is 00:0E:A6:6B:DD:C3)
Network B (GIT PC is 00:11:2F:3F:29:AC)
Network C (GIT PC is 00:0E:A6:46:19:6A)

Network A Gateway MAC is 00:0D:88:2F:A8:B7
Network B Gateway MAC is00:0F:66:41:7B:D8
Network C Gateway MAC is 00:0C:41:A3:20:F6
Viatrak
 
Posts: 5
Joined: Thu Dec 30, 2004 6:23 pm

Postby Ark » Thu Dec 30, 2004 10:05 pm

Those results look just fine.
To annonate them a bit more:

LAN 1
00:0E:A6:46:19:6A - 25 C
00:0F:66:B7:06:5C - 25 (extra B)
00:11:2F:3F:29:AC - 25 B
00:02:E3:23:D8:DF - 25 C
00:0F:66:41:7B:D8 - 22 B-GW
00:0C:41:A3:20:F6 - 9 C-GW
00:0C:41:5F:0D:CA - 10 (extra C)

LAN 2
00:0C:41:A3:20:F6 - 27 C-GW
00:04:23:59:90:EA - 30 A
00:02:E3:23:D8:DF - 30 C
00:0E:A6:6B:DD:C3 - 30 A
00:0E:A6:46:19:6A - 30 C
00:0D:88:2F:A8:B7 - 28 A-GW
00:0C:41:5F:0D:CA - 14 (extra C)

LAN 3
00:0F:66:B7:06:5C - 9 (extra B)
00:0E:A6:6B:DD:C3 - 30 A
00:0D:88:2F:A8:B7 - 8 A-GW
00:04:23:59:90:EA - 30 A
00:0F:66:41:7B:D8 - 13 B-GW
00:11:2F:3F:29:AC - 30 B

Clearly no LAN is receiving traffic that was originally generated on its own LAN, so nothing is being duplicated or sent in a circle. There is obviously a 2nd computer on LAN B and a 3rd computer on LAN C which are not trying to play the game but are getting some amount of traffic tunneled (even if its just ARP traffic, hence the time until they expire is lower, since that time resests to 30 seconds every time GIT receives a packet that was from that MAC and ARP traffic is pretty few and far between)

The way GIT works is that all traffic generated on network A will be sent to network B and network C, and neither B or C will reproduce that packet back to any other network. Keep in mind that your upstream bandwidth becomes very important now and a slow 128K cable modem upstream speed on network B means you only have 32K per computer if it has to be sending separate packets out to each player on network A and C.
Ark
Site Admin
 
Posts: 2108
Joined: Sat Sep 13, 2003 4:21 pm

Postby Viatrak » Thu Dec 30, 2004 10:37 pm

The way GIT works is that all traffic generated on network A will be sent to network B and network C, and neither B or C will reproduce that packet back to any other network.



--

By this, do you mean traffic generated on Network A for Networks B and c would be "merged to form traffic BC" sent to B and C locations; instead of sending B to B and C to C?

If you mean sending BC to A and B seperately is there a way to send B to B and C to C instead (Through GIT?)?

Also, would changing the reset time of 30 seconds to anything lower (If possible somehow....) change the storm?





Finally, the "storm" occurs in the game lobby with 5 players present. It doesn't occur with 4 players (1 Player less from networks A or C). It occurs in the lobby even when no chatting occurs. When one player leaves it returns to norm. Anyway, this game is StarCraft and it doesn't seem to be a game that takes much bandwith...




(Note: Sorry if some of this doesn't make sense, some are just truly "random ideas")
Viatrak
 
Posts: 5
Joined: Thu Dec 30, 2004 6:23 pm

Postby Ark » Fri Dec 31, 2004 12:09 am

I mean that traffic generated on network A that is going to be tunneled according to your options set, will be tunneled to both B and C no matter what, since you have two connections set up in GIT (port 213 and port 214).
The same data it sent out on each port. GIT has no way of knowing if certain data is only meant for network B or C.

Changing the 30 second timeout would not change anything useful. Setting this value to a lower number as far down as 1 second would likely have the same results. Setting it really low like 0 or a fraction of a second if possible, would only result in GIT duplicating traffic such as sending back a packet that was originally generated on network A and sent to network B, by having the GIT on network B forget that the packet came from network A, and send it back. Setting the value any higher then 30 is just going to waste memory if you have a large network with 100s or more computers or something.
Ark
Site Admin
 
Posts: 2108
Joined: Sat Sep 13, 2003 4:21 pm


Return to GIT

Who is online

Users browsing this forum: No registered users and 22 guests

cron