Info regarding "WAN dies"

[Cantara]

I opened a new thread to discuss this because the original thread is already too big and confusing.

For almost Emule/Edonkey users had already noticed that sometimes your internet connection dies without a 'logic' explanation. The connection simply dies, and after a while (or when you make release/renew on you router) the connection is again established.
By reading the thread, I think I can suppose that this happens when we have a high number of simultaneous connections, and the router simply can't handle them.

I was reading the EmulePlus FAQ, when I found this important information:

Sleepy's router guide v0.3

The following is about how to set up eMule itself when using a router, this is important so do not ignore it.

Most routers use a 16 bit address code for calculating the connections, this results in a maximum amount of connections of 256, normally this would not be a problem. But when using eMule it can be a problem, since eMule usually requires as much connections as it can get.

The max connections can be set in the connection tab in the preferences menu, without a router this can be set to about 300 to 500 on a windows 2000/XP system (less than 90 on a windows 95/98/ME system because of the bad networking system in these versions).
A router uses a special chip to arrange connections, this chip has limitations!
When using a router you have to keep an eye on this, I advise setting this to around 200 and not very much higher because if the router has to many connections it can crash(will need a reboot then) or stop making new connections. This means no more web browsing or getting email and things like that because these connections can not be made by the router.

Most routers suffer from this problem.

Examples; Linksys routers, Alcatel routers (especially the 510i and the speedtouch PRO versions) Netgear routers, some cable modems even suffer from this problem.

When you set the max. connections lower, to 200 do not set max sources per file to high, otherwise it will keep looking for sources and forget to keep an eye on sources where you can download from. Set these to 400 when downloading 5 or less files, to 300 when downloading 5 to 12 files and with more then 12 files set it to 250 or even 200. These settings can be found in the Files tab in the preferences window.
eMulePlus is a very smart little mule, very uncommon for mules but it comes in handy, if you can get more sources than those that you already have emule will discard the useless sources slowly and let fresh sources in.

The last setting to set is the Max. connections per 5 seconds, routers with a faster processor(yes every router has it's own processor) can handle more connections per 5 seconds, I recommend setting this to 30. If you think your router can handle more, then gradually increase this value until you notice slower internet and failing to load web pages.

So, if this information is correct, the Asus WL500g is OK and the problem consists in too many connections at the same time. If you are an Emule/EmulePlus/Edonkey.... use this tips to configure the program to not overload your connection!

I hope this info helps you!

[Snigel]

Too bad that this issue also is when high speed is going through wan with one connection :/

[Antiloop]

what is high speed in your opinion?
i can receive without problems 4gb (saved to the usb harddisk directly) at my line

but i only have a 2,5MBit connection for download..

[Snigel]

Well, I have things a bit strange connected, so it's actually both internet and lan from some other computer coming through wan.

My internet is 10mbit, that's not the real problem. But it dies really fast if I try to send something at 100mbit.

My connection looks like this:

Server -> Router ->
-------------------------> Switch -> Adsl-modem
Laptop+Workstation ->

I don't know if my internet can kill the connection, some guy here said earlier in this thread that 10mbit could kill his line, just that it took longer time than 100mbit.

[Oleg]

Do you send it using wired LAN or WLAN? Once your WAN port "dies" do you've an access to wl500g?

[Snigel]

Yup, my wl500g is connected via the wanport to a switch. So it accesses both the rest of my network and internet throug wan.

[SocoNaTromba]

I've recently bought a wl500g router, and dsl-300t d-link adsl modem.
I was suffering from this "wan port dies" problem... tried lot of cofigurations in the modem and in the router... but it kept crashing upon large amount of traffic, although the lan conections still worked perfectly, and I was even able to telnet to the router ..., everytime it crashed I used to hard reboot de router to get back internet connection again.
This used to happen even after I updated de router with the latest Oleg's firmware ( thankz oleg for your great work and dedication, it should be recognized "finantially" by asus)...

... I was really considering returning the router to the seller, till I tried the following:

I configured modem with the ip 192.168.0.254. then I configured it to work in bridge mode, and its dhcp server disabled. then I configured the router with the lan ip 192.168.0.110.

The "speed negotiation method" seem not to work with this modem, as its ethernet led never lits up when I set negotiation other then "auto".

I've configured manualy the DNS server 1 , as the second is locked to router's ip.

...configured username and password for dsl connection....

putted the wireless channel fixed to channel 6. changed SSID...
...blanked router's host name....

mac filtered to the pc's on the lan and to 2 wireless cards...
activated dhcp and manual ip for each mac address inserted.
distinct port forward to both lan connected computers so azureus could accept remote connections.

every other config as defaults...

well well, what did I do ?!

Now the connection IS completely stable, with constant heavy trafic, using 2 computers on lan both with Azureus connected 24H/day using all the bandwith, with frequently 400KB/s download speed / 100KB/s up speed ,,, and its efficiency is completely ok 100% without crashes... just the forced disconnection every 18 hours imposed by the isp...wich reconnects in aprox 30secs...

And the really funny thing:
when I try to, after all this configs, open http://192.168.1.1/ it prompts me for username and password, indicating: "level 15 access"... admin/admin does not work...

remember that my configs uses the group 192.168.0 instead of 192.168.1

it shouldn't exist a 192.168.1.1... on my local network...

so I made a trace route to that ip...

C:\Documents and Settings\Hugo Pinto>tracert 192.168.1.1

Rastreio da rota para 192.168.1.1 até o máximo de 30 saltos

1 <1 ms <1 ms 1 ms 192.168.0.110
2 18 ms 19 ms 19 ms r6613.nr.ip.pt [195.23.0.154]
3 19 ms 19 ms 19 ms 195-23-31-126.nr.ip.pt [195.23.31.126]
4 20 ms 22 ms 139 ms 195-23-31-129.nr.ip.pt [195.23.31.129]
5 52 ms 23 ms 27 ms 194-79-92-9.nr.ip.pt [194.79.92.9]
6 60 ms 25 ms 26 ms 195-23-31-234.nr.ip.pt [195.23.31.234]
7 27 ms 23 ms 25 ms 213-205-74-41.nr.ip.pt [213.205.74.41]
8 52 ms 31 ms 28 ms 192.168.1.1

Rastreio concluído.

hum? 192.168.1.1 goin out to the adsl connection?
could this be the reason of the wan crashes?

Because if somebody uses 192.168.1.1 as lan ip for router, the default ip for router, I wonder if it is somehow conflicting with the other ip existing on the internal network of my ISP...

I read somewhere at chupa.nl that the "wan dies" problem has something to do with the dns requests... so thats why I configured it manually...

Could any body that has the "wan dies problem" checkout if a tracert 192.168.1.1 , after configuring the ips with the 192.168.0 class, returns any jump?

Best regards

Hugo Pinto

[Nominus]

I have found some kind of solution - add the folowing line
/bin/watch -n 60 /usr/sbin/et -i eth1 speed auto &
into the /usr/local/init/post-boot and do flashfs save, flashfs commit, flashfs enable (as it's described at http://wl500g.dyndns.org) and reboot the router. After that every 60 seconds your router will test the connection for speed rate and will restore it.....

Or better you can make a script like this:
#!/bin/sh
ping -q -c 1 your_gate_IP > /dev/null
if [ $? -ne 0 ]; then
/usr/sbin/et -i eth1 speed auto
fi
place it in /usr/local/bin and watch it instead of "/usr/sbin/et -i eth1 speed auto" directly.

I know this is very ugly and unnatural solution but it works...

[Oleg]

So, could you confirm, that knocking et with "speed auto" fix the problem?

[Nominus]

So, could you confirm, that knocking et with "speed auto" fix the problem?

Yes, I could confirm that it solves the problem of dying WAN port during high speed (3-4 Mb/s) FTP transfer... at least WAN port restores quick enouth. But the better way is not to knock, but to test (ping) connection before do "et -i eth1 speed auto" as i described earlier.

[Oleg]

ok. we've made similar findings with Antiloop.
Can you please check that

ifconfig eth1 w.x.y.z

also helps (w.x.y.z is the current IP address, assigned to wl500g, including 0.0.0.0 if you use pppoe).
"speed auto" thing is doing to much, it's to hard to isolate problem.

[Oleg]

And another thing - try just this instead of speed auto, once your port is dead

et -i eth1 up

and if it does not help, try

et -i eth1 speed auto

Post results here.
Also, I need results of command below before & after each et command (wait a bit after each command execution, so things stabilize)

for a in 0 1 2 3 4 5 6 7 16 17 18 19 20 21 22 23 24 28 29 30; do echo -n "$a:$(et -i eth1 phyrd $a) "; done; echo

Everyone is invited.
So, the sequence should be as follows. Once you port died, telnet to router and copy/paste these lines

nslookup www.google.com
for a in 0 1 2 3 4 5 6 7 16 17 18 19 20 21 22 23 24 28 29 30; do echo -n "$a:$(et -i eth1 phyrd $a) "; done; echo
et -i eth1 up
sleep 2s
nslookup www.google.com
sleep 5s
for a in 0 1 2 3 4 5 6 7 16 17 18 19 20 21 22 23 24 28 29 30; do echo -n "$a:$(et -i eth1 phyrd $a) "; done; echo

Copy the results here. Check your connection - if it's alive, don't do anything, otherwise copy/paste this

et -i eth1 speed auto
sleep 2s
nslookup www.google.com
sleep 5s
for a in 0 1 2 3 4 5 6 7 16 17 18 19 20 21 22 23 24 28 29 30; do echo -n "$a:$(et -i eth1 phyrd $a) "; done; echo

Check your connection again.

[Nominus]

Yes, `ifconfig eth1 10.0.2.7 broadcast 10.0.2.255` restores WAN port, but it breaks default gateway record in the routing table, so i have to do `route add default gw 10.0.2.1` to restore it (10.0.2.7 - WAN IP, 10.0.2.1 - gate's IP).

btw, i've noticed that the router sets broadcast adress wrong - 10.255.255.255 instead of 10.0.2.255 (my mask is 255.255.255.0) and i have found no way to set it properly in the Web-interface, only by typing`ifconfig eth1 10.0.2.7 broadcast 10.0.2.255 mask 255.255.255.0` directly.

and i did as you had asked:

~ # nslookup www.google.com
*** Unknown host

nslookup: www.google.com: Unknown host
~ # for a in 0 1 2 3 4 5 6 7 16 17 18 19 20 21 22 23 24 28 29 30; do echo -n "$a:$(et -i eth1 phyrd $a) "; done; echo
0:0x3000 1:0x786d 2:0x0022 3:0x5521 4:0x01e1 5:0x45e1 6:0x0007 7:0x2801 16:0x1800 17:0x006d 18:0x0f90 19:0x0000 20:0xc0b0 21:0x0003 22:0x0fdf 23:0x8001 24:0xb65c 28:0x0044 29:0x1000 30:0x3a60
~ # et -i eth1 up
~ # sleep 2s
~ # nslookup www.google.com
*** Unknown host

nslookup: www.google.com: Unknown host
~ # sleep 5s
~ # for a in 0 1 2 3 4 5 6 7 16 17 18 19 20 21 22 23 24 28 29 30; do echo -n "$a:$(et -i eth1 phyrd $a) "; done; echo
0:0x3000 1:0x786d 2:0x0022 3:0x5521 4:0x01e1 5:0x45e1 6:0x0005 7:0x2801 16:0x1800 17:0x0000 18:0x0e99 19:0x0000 20:0xc0b0 21:0x0003 22:0x0fdf 23:0x8001 24:0x4abf 28:0x0044 29:0x1000 30:0x3a60
~ # et -i eth1 speed auto
~ # sleep 2s
~ # nslookup www.google.com
Server: gate2.maryno.yes
Address: 10.0.2.1

Name: www.google.akadns.net
Address: 216.239.59.147
~ # sleep 5s
~ # for a in 0 1 2 3 4 5 6 7 16 17 18 19 20 21 22 23 24 28 29 30; do echo -n "$a:$(et -i eth1 phyrd $a) "; done; echo
0:0x3000 1:0x786d 2:0x0022 3:0x5521 4:0x01e1 5:0x45e1 6:0x0007 7:0x2801 16:0x1800 17:0x006d 18:0x0f90 19:0x0000 20:0xc0b0 21:0x0002 22:0x0fdf 23:0x8001 24:0xb65c 28:0x0044 29:0x1000 30:0x3a60
~ #

I think nslookup isn't a good choice to test the connection, cos timeout is *TOO* long, imho `ping -c 1 host` is much better.

[Oleg]

So, could you please confirm, that et -i eth1 up does not help?

[Nominus]

Yes, i've tested it again and `et -i eth1 up` steal does NOT help.