AT&T, 2Wire, and Disconnections
by Arandur | 25/11/2007 08:53:25
I have been struggling with an issue for a little while now that I cannot explain.
When I perform certain actions in-game, my 2Wire modem/router loses sync and drops connection.
It always happens in battlegrounds, it always happens in raids, it also always happens when I perform a crafting action by using "create all." In battlegrounds and raids it happens within ten minutes. It also seems to happen when I'm in a heavily populated area like IF at peak times.
When I get disconnected the first time the modem drops sync. The modem quickly reconnects and I log back in. Then zone loading screen comes up, but as soon as the bar reaches the end and the UI comes up, the modem loses sync again and I am disconnected. This happens over and over until I give up and wait an hour or so. This happens consistently and without exception.
Here is my equipment:
I have a 2Ghz MacBook running OSX 10.4.11
My modem/router is a 2701HG-B Gateway and my ISP is AT&T DSL.
All Firmware is up-to-date.
Here is the long list of things I have tried and narrowed down:
- It never disconnects when using the network at my school.
- I never lose my wireless connection, the modem itself loses sync to the DSL.
- I have deleted my Addons/Cache/WTF folders.
- I have tried a clean install of Warcraft and TBC with no addons.
- I have tried a clean install of the OS and then another clean install of Warcraft.
- I have tried all of the port-forwarding on the router recommended in the support forums.
- AT&T as been to my house and checked the line, there are no problems on the line.
- AT&T replaced my gateway with a 1701HG Gateway and there was no change.
- AT&T did a "rip and rebuild" on my DSL with no change.
- The AT&T guy came and watched me play three battlegrounds with the exact same problem happening on all of them at about the same time (right around 10min in). He also personally witnessed the modem losing sync EVERY time I tried to reconnect after that, at the exact time that the loading bar finishes.
- Needless to say, he was stumped.
- AT&T currently has no idea what to do about this.
- The modem never disconnects when I am just browsing, or watching videos online, or downloading, or anything else. Only when I play Warcraft, and specifically when I am in a BG, Raid, or a few other things like in IF with tons of people around.
- I am currently visiting my GF at her grandparent's house for Thanksgiving. Her grandparents happen to also have AT&T DSL with a 2701HG-B Gateway - the exact same problem happens here! This is two States away in a different time zone.
On a positive note, I've finally gotten around to leveling my fishing skill since that appears to be the only thing I can do when I log in. (370, up from about 86!)
Obviously this is very upsetting and frustrating.
How can actions in game cause my modem to lose sync?
The AT&T technician certainly doesn't think it is possible (or didn't until he saw it himself).
What in the world is happening?
by Tigerclaw | 26/11/2007 00:59:53
This is interesting info. One thing we might consider experimenting with is forcing a smaller MTU size.
The default MTU setting is usually 1490 or 1500 bytes, it might be interesting to change your MTU down to 1400 as a test and see if the modem has fewer re-training incidents.
To me re-training implies that the modem has detected an unacceptably high number of data transmission errors or re-tries on the DSL link and is trying to re-calibrate the link to get more reliable data flow again. It might be the case that forcing the link to use shorter packets could relieve this issue to some extent. This is strictly a guess on my part.
by Tigerclaw | 26/11/2007 03:08:47
Network protocols (and more importantly the exact content of packets sent) can change from patch to patch, from player to player, and from character to character. There are no sane steps Blizzard can take to avoid this.
A router and/or modem needs to be able to pass any requested network traffic reliably or it will result in frustrated users.
by Tigerclaw | 26/11/2007 03:47:36
"DSL Retrain" sounds like the modem had trouble sending a certain bit pattern over the wire and is trying to reconfigure the link in vain. You might ask if your DSL provider can opt to deliberately run the link at a lower speed than what they believe is the peak possible speed. It's another guess that might yield some results.
You didn't mention (or I missed mention of) whather you have tried playing on ethernet cable instead of wireless. Might be wrth a try as a diagnostic aid if nothing else.
I'm also thinking you might want to try and set the MTU to a lower number on the 2Wire itself, this may require learning how to log on to that modem's config page and adjusting it. It's possible that changing it on your computer may not have the desired effect. For all we know it has to do with some bit pattern coming down from the server (perhaps big packets or certain packet types) that is not being handled well by the 2Wire in its default config.
Trela, if you can send me any more detailed references on this type of issue with 2Wire hardware, we'd appreciate it. ->email@example.com
by Tigerclaw | 26/11/2007 04:45:03
There must be something about those particular bit patterns that is confusing the router.
You've gathered a lot of good test cases here, the next step is figuring out how to get the attention of 2Wire and AT&T focused on this bug so they can identify and fix it.
by Tigerclaw | 28/11/2007 18:01:50
Stickying this thread for a while to see how the reports pile up.
by Tigerclaw | 29/11/2007 01:52:28
This situation has appeared in one form or another over several years.
Each WoW patch makes changes in packet content. These are just strings of bits that we send through the OS network facilities.
If a modem or router locks up or loses sync or connection when it tries to send certain bit patterns, that's a bug all right, but it's not in WoW.
by Tigerclaw | 29/11/2007 20:31:03
Are you also experiencing the situation where the router must be re-set, not the computer ?
by Tigerclaw | 29/11/2007 20:33:34
That's a good question (is it OK to trace packets in order to troubleshoot).
My personal POV is that we seem to have zeroed in on a few specific models of 2Wire router that are problematic. If I was in your shoes I would try these things first:
a) ask your DSL provider to lower your connection speed a few notches to see if it starts behaving.
b) think about replacing that modem.
c) it wasn't clear from the post but when at home, do you also use wireless to get form the laptop to the modem ? Try ethernet if so since we have been made aware of specific bugs in 2Wire's that crash the router when used with certain flavors of wireless-chip in the laptop.
This type of issue is just not fixable from the application software side.
by Tigerclaw | 30/11/2007 04:50:32
Are you using wireless or ethernet, and have you tried changing from one to the other ?
by Tigerclaw | 30/11/2007 06:33:04
The smart thing for all such affected owners to do, is make sure 2Wire knows about this problem and how many people it affects. There is not much we can do here about it and I doubt their staff are reading this forum.
by Matthewjp | 30/11/2007 22:48:54
I wanted to provide an update to this issue. We have been in contact with 2Wire about this issue and their suggestion is to avoid downloading anything while playing. Tigerclaw's earlier suggestion to contact your DSL provider and have them lower your connection speed can help with this.
2Wire has also requested that users contact 2Wire directly and provide any DSL statistics they have. You can contact 2Wire through the following site:
Click on the My Stuff link and create an account with them if you don't already have one. I'd suggest referencing this thread in the ticket that you submit.
by Matthewjp | 03/12/2007 21:17:08
We received direct contact from 2Wire on this issue. From the information we received, 2Wire is suggesting that any user with this problem contact them directly. Please continue to do so by using the instructions I posted earlier:
AT&T, 2Wire, and Disconnections
If we receive any further updates, we'll post them.
by Tigerclaw | 13/12/2007 00:02:06
If 2Wire can present us with a clearly defined rule showing which types of packet content will or will not crash their router, we might be able to dodge that bug. It would be a hard sell to the development team though, putting in a workaround for a specific misbehaving router firmware revision.
Or, if 2Wire wants to let us know in some detail where they feel the game is violating any rules of TCP/IP protocol, we could look at it from that angle. However the evidence favors an interpretation that the game is in-spec and the router isn't. Which is another way of saying, posts on their support forum might be a better way to get their attention on this class of issue.
by Tigerclaw | 20/12/2007 19:46:21
Has AT&T offered any means by which they might run the line at a lowered bit rate to see if it is some unique kind of signal quality problem ?
by Matthewjp | 24/12/2007 22:33:43
I've received word from another World of Warcraft player that their issue was resolved when AT&T replaced their Speedstream 4100 modem with the Motorola Netopia 2210. Here is Motorola's information on this modem:
Please use that link as a reference when speaking with your ISP. Please ask if they can try this modem on your connection instead of the Speedstream.
by Tigerclaw | 25/12/2007 01:08:10
Best news I have seen on this topic so far. Thanks for posting it.
by Tigerclaw | 25/12/2007 04:52:15
Correlation is not the same thing as cause and effect.
This is a class of situation we have seen many times since the launch of WoW. Some routers have bugs. When WoW's use of the network changes, sometimes bugs that were not previously noticed can begin triggering. If you see behavior where your router actually needs to be reset, this is well outside of our ability to fix. Even assuming we could start to guess at which traffic patterns take your modem or router offline, WoW doesn't actually control the packet sizes being sent through the link, the OS does. So it's just not a road we can even begin to drive down. The router needs to be fixed or replaced.
At one time the problem child was the Linksys WRT54G - its behavior would change drastically from one patch to the next, and the same sorts of discussions ensued - "it didn't act up until WoW changed, so just un-do whatever you did in WoW and it will be OK" - eventually a firmware update appeared for those modems and they became more reliable.
The best thing you can do is bring this misbehavior to your ISP's attention. There shouldn't be any circumstances where *any* particular burst of traffic between server and client should cause a router to hang up or drop connection.
by Tigerclaw | 26/12/2007 18:10:20
On the other hand we have reports of users who have traded in their Speedstream modem for a different model and WoW began behaving normally again. WoW isn't aware of what hardware it is using to pass the network traffic and it hasn't been patched - the only logical conclusion in those cases was that their original modem was not 100% reliable and the new one was more so.
This is a classic interaction problem. You're right to surmise that the interaction might cease if WoW's network traffic patterns change again. Whether it will be for the better or worse on a given modem is difficult to predict in the presence of bugs, since we as yet have no documentation of what kinds of bugs are present in the affected devices.
by Matthewjp | 27/12/2007 18:08:24
That's great news. Thanks for posting. Could you inform everyone of which modem you were upgraded to? That will help others with modem suggestions.
by Tigerclaw | 29/12/2007 02:07:09
If you have one of the misbehaving modems, you don't necessarily need to replace it. You could instead ask to get a firmware update from the manufacturer to fix whatever bug it has, assuming such an update is available.
Firmware update from manufacturer: available whenever they decide to do it
Replace modem: available whenever you decide to do it.
The best thing for affected users to do is being this to their modem maker's attention directly.
by Tigerclaw | 31/12/2007 02:51:15
Thanks for this report. Where and how did you adjust the MTU size in the router ?
by Tigerclaw | 01/01/2008 06:16:23
WoW merely opens a connection and sends and receives chunks of bytes through it. If there is a bug or interoperability problem with any particular wireless hardware or router vs. a particular class of system or OS - it is not a problem that changes to WoW can solve conclusively.
Looked at from another angle, if WoW was doing anything that could conceivably cause illegal TCP packets to be sent, this would be the fault of the OS since WoW does not control what packets are put through the linking hardware - the OS does. More to the point if WoW was doing such a thing, it would have negative effects across all known internet routers. It just wouldn't work anywhere.
We have seen this situation in the past (example, Linksys WRT54 series) and the conclusive solution has always come from the router or NIC manufacturer with updated firmware or drivers (read that as "bugs found and fixed").
by Tigerclaw | 01/01/2008 19:29:54
See post 120.
by Tigerclaw | 02/01/2008 00:10:22
This turned out not to be the case for the WRT54G. Other applications also worked fine, and WoW started working fine once the firmware was updated.
Bugs in routers may be data rate, packet size or packet content specific.
Another way to look at this is - you could put a laptop in a room with ten different makes of router and find WoW working on 9 out of 10 of them - sending the exact same forms of data through each one. But when you get to #10 the game doesn't work and the router hangs up. Where would you think the bug was ?
by Tigerclaw | 02/01/2008 04:32:53
One often overlooked fact is that there are multiple brands/makes of modem using the same reference design hardware and firmware out there. This has been documented on some of the router hacking & enthusiast sites.
The reality is that the vast majority of routers work correctly.
by Tigerclaw | 03/01/2008 01:32:25
Or, as was the case with the WRT54G, having a conversation with your modem vendor about a firmware update. The more people notify them about the interoperability issue, the more importance it will be assigned.
Or, as seems to be the case with some posters here, try some config changes to the unmodified router to see if you can work around the problem. I saw one user who set a smaller MTU value in the router config and got better results. I saw another user who disabled 802.11g for a wireless station and went back to 802.11b and got better results. There may be more such changes you can try.
by Tigerclaw | 03/01/2008 20:06:49
As predicted, changes in network traffic can cause router bugs to become active or dormant.
by Tigerclaw | 03/01/2008 23:50:26
I guess we're just going to have to agree to disagree on that point. Taking a broader POV, we have done a modest amount of research on the affected router makes and models, and even in that limited amount of Googling we found plenty of evidence to suggest that WoW is not the only application or usage pattern that runs into trouble on some of these modems and routers.
by Tigerclaw | 31/01/2008 19:49:15
If connecting wirelessly to your 2Wire, try re-configuring the 2Wire for 802.11b mode (11MBit) instead of the 802.11g mode (54MBit). Or, try connecting by ethernet cable instead of wireless.
by Tigerclaw | 13/02/2008 01:31:06
I hear your pain but you didn't actually post what modem you are using.
As others have established, there are some makes and models of DSL modem and router that are problematic (unreliable when faced with certain forms of traffic which work fine on other routers).
There isn't something we can do to fix that class of problem once and for all, a router that chokes on one kind of data pattern might well choke on thousands of unique patterns - no way for us to predict which ones are at fault and realistically, no good reason why we should special case networking code to work around modem/router bugs.
by Tigerclaw | 23/03/2008 02:40:24
What this means is that there is a bug in your modem which doesn't properly handle all data patterns. New patches introduce new data patterns (and as you can see it is context sensitive). We have no idea which particular set of packets causes their modem to crash - but we know that other modems don't in the same situation.
A better audience for this type of report might be 2Wire's support forum - so they are made plainly aware of the bug in their modem, perhaps they can come up with a firmware update to eliminate this behavior.
Or, you could try a different make/model of modem as has also been posted on this thread.
by Tigerclaw | 27/03/2008 15:27:01
I won't delete your post but I will stand by the statement that modem/router firmware bugs exist, firmware updates have resolved modem/router lockups in some cases, and switching modem make/model has also resolved lockups in some cases (read this thread for example).
There isn't anything we can easily change in WoW to prevent buggy modems locking up unless the vendors want to provide a list of which data patterns can cause their device to fail - it is better to focus attention on making vendors aware that there is a problem, so they can test it and fix it.
by Matthewjp | 28/03/2008 15:28:53
I'd just like to add that I've personally troubleshot this issue since early 2005 on Windows & Macs. The first piece of hardware that we started seeing this issue with was the Linksys WRT54G(S) router. All a user had to do back then was bypass the router and it was resolved. There was a 3rd-party firmware that someone put out in April 2005 that resolved it and Linksys put one out in July 2005 that resolved it as well.
I don't believe this issue has become worse. Using this same piece of equipment as an example, the newer revisions of it, I believe starting with version 5, don't have this issue. In fact, it's been a while since I've seen this router cause this issue. Nothing changed on our end to resolve it. A firmware update and a hardware change fixed it.
I've personally fixed this issue by changing network card settings on Windows systems as well. Replacing hardware is no where near the number one way to resolve it. I hope you decide to take the appropriate action necessary to resolve this issue.
by Tigerclaw | 03/04/2008 23:21:49
Different OSes (and drivers and NICs) have different characteristics in terms of the bit patterns put on the wire or wireless. The data you are supplying doesn't contradict our premise - that some routers and modems are vulnerable to specific type of bit patterns at the packet level.
It could be something as simple as the MTU chosen, to possibly something like a timing-sensitive bug which would in turn be dependent on the timing parameters of the TCP stack in question.
By way of example, we saw a rash of this class of problem as users upgraded individual machines form Tiger to Leopard. No hardware changes, just a change in the OS and TCP stack, and router problems surfaced.
Bumped: The Impending Cataclysm