Netgear DGN2200 doesn't like routing only?

[#1]

Originally posted on the Netgear Support forums but went without response so it's over to you guys
I have used my DGN2200 as the sole router and ADSL Max (UK, G.DMT) access point, this worked perfectly for both wired and wireless connections.
I've now changed to using a DG834G as the ADSL access point and the DGN2200 as a switch / wireless access point.
To do this I changed the following:
*On the DG834G I:
-Turned wireless off
-Physically removed the aerial
-Changed the DHCP range to x.x.x.10 - x.x.x.254
-Set up all my usual ADSL and port forwarding settings
*On the DGN2200 I:
-Turned NAT and firewall off
-Changed router IP to x.x.x.2
-Turned off DHCP server
This network configuration is needed so that I can:
-Network all rooms in my house (cables are now in the walls)
-Have wireless access throughout the house and garden
-Have the ADSL access point next to the master socket
The problem:
Wireless access keeps dropping. I've tested with 2 PC's (one built in Dell wireless, one Belkin PCI card) and an iPad. The devices can see the network but 'obtain' the default 169.x.x.x IP address. This can only be resolved by rebooting the DGN2200. This happens to devices that were already connected to the network, and those trying to join it that weren't previously connected.
During this time, devices hard wired to the DG834G have no connectivity issues whatsoever.
I'm going to turn wireless back on on the DG834G to provide a quick and easy test to see if the issue is there, but I doubt it due to the previous checks and resolution of rebooting the DGN2200. I also have other routers (next one to test is a WNR1000) but I'd prefer to use the DGN2200 as this is the only 40MHz device I own.
Does anyone have any advise as to what might be causing the problem and what to do to resolve it?
Thanks in advance.

Since I wrote the above post I have turned DHCP 'on' on the DGN2200 in a different range (same subnet) and reduced the DHCP range in the DG834G accordingly. The wife has had similar no connectivity problems but I've not been home to test if it's the same problem, it resolved itself after a minute or so.

[#2]

Hi Ben,
During these periods of new clients not being able to connect to the network (and defaulting to the 169 addresses) what happens to those clients already connected i.e. those that are not in need of DHCP support? Do they still remain connected?
If you haven't done so already, I'd try temporarily disabling all DHCP and configuring static address on everything and observing things for a while to see if clients can remain fully connected. If they can then this obviously points to a DHCP issue (perhaps you've already determined this - I wasn't sure from your message if that was the focus).
It would also be worthwhile running a packet sniffer (e.g. Wireshark) on a wired client and observing what packets are floating around during periods when a new client attempts to join the network but fails. You should be seeing DHCP broadcast packets - if you are not then this points to the DGN2200 not flooding such packets from the wireless side of its bridge to the wired side (and on to the DG834G). If you are seeing these packets then you should also be seeing responses being broadcast (but from the wired clients viewpoint you wouldn't know if they are making it over to the wireless side of the bridge).
I wouldn't run with multiple DHCP services, even if they are using different pools, as you could end up with a bit of a muddle at best whose problems might manifest themselves in even more mysterious ways!
Mathew
P.S. You mention in your subject that the suspicion that the DGN2200 doesn't like 'routing only' however note that in this scenario it's not routing - it is behaving purely as a layer 2 device insofar that it is simply bridging a wireless and wired network segment together - they are both operating within the same IP subnet hence there's no routing between them, merely targetted layer 2 frames (based on MAC addresses) and broadcasts (or not as the case may be here) being passed over the bridge.

[#3]

What's the lease time on the DHCP?  Are your clients getting a 169.* address on renewal only?  Try setting a much longer lease and seeing if the problem recedes accordingly.
I was going to make some other suggestions by MJN beat me to it, if you can I'd also see if another DHCP sever on the network has the same problem.

[#4]

@MJN
Great advice, thanks. I'll revert to the single DHCP server and re-test when connectivity goes down on one machine and report back full results.
@fourfourdevon
I can't actually find a DHCP lease time in the DG834G but I don't think this is the problem as it will affect two wireless devices at the same time.
I think DHCP is a bit of a red herring and the 169 address is obtained simply due to a lack of connectivity between the DGN2200 and the DG834G. I think the DGN2200 doesn't like not being the sole access point and throws it's toys out of the pram after running for a while.
Edit: All of the built in diagnostics for the DGN2200 show no external connectivity, despite clients currently working OK. I need some way to turn this router into a dumb wireless access point!

[#5]

Quote from: Ben

I think DHCP is a bit of a red herring and the 169 address is obtained simply due to a lack of connectivity between the DGN2200 and the DG834G.

I'd say the DHCP issue is the most important piece of evidence we've got!  If DHCP fails then the TCP/IP stack will fall back to a randomly generated 169. address as an auto-configuration mechanism. Solving why the DHCP is failing is key to solving the problem in my opinion.

Quote

Edit: All of the built in diagnostics for the DGN2200 show no external connectivity, despite clients currently working OK. I need some way to turn this router into a dumb wireless access point!

That's to be expected and entirely consistent with its role here. From the DGN2200's perspective there is indeed no external (WAN) connectivity but that's not what you're using if for in this scenario - it is being used purely as a switch to bridge wireless and wired networks together to form a single LAN. Hence, all your clients - including your Internet gateway - should thus be able to see and talk to each other. Your external connectivity is obtained via the DG834 which your DGN2200 really needs no explicit knowledge of (other than knowing its MAC address and which switch port it is connected to).
Mathew

[#6]

Quote

I think the DGN2200 doesn't like not being the sole access point and throws it's toys out of the pram after running for a while.

As MJN says, in this configuration the only functionality that you're using on the DGN is the switch. Maybe it doesn't like the fact that its not been able to establish a WAN connection for some time and goes into 'sulk' mode as you say.
When its in this situation can you get to its Internal IP x.x.x.2 ( and maybe web configuration? )  from anything directly connected to the DG834 ?
If so is there anything useful in the log which might give an indication of what its problem is ?

[#7]

Hi
Why dont you just use the dgn2200 as the router/adsl/wireless and dg834g in WDS repeater mode? With DHCP off etc. If you log-in into the dgn2200 you will see a Wireless Repeater Fuction. This works really well to give you 100% wireless n
James

[#8]

@james26
I'd like to keep some devices hard wired and avoid the losses in wireless performance with repeating. In addition, I need the DGN2200 in the middle of the house to shunt wireless across my 4 stories and the garden which it can't do if placed by the master socket and the DG834G can't get wireless to the bottom of the garden no matter where it as, as it's 2.4GHz -G instead of 5GHz -N.
@Everyone else
I'm just waiting for the problem to re-occur so I can check all the suggestions provided.

[#9]

Quote from: Ben

@Everyone else I'm just waiting for the problem to re-occur so I can check all the suggestions provided.

It may well be the case that the unrepeatability is the nub of the problem, and could therefore prove impossible to fix. Residential routers are made down to a price and hence all too often the software stacks can be quite buggy when it comes to the nuances of proper operation. In most cases such problems can go unnoticed (by virtue of higher level error detection/correction techniques such as TCP) however when it comes to DHCP there's a finite window of time within which it can work and if this fails there're no second chances.
Mathew

[#10]

It seems to happen once every 2-3 days, I've just been using my main computer (hard wired to the DG834G) recently instead of my iPad so I've not been able to run the checks advised when the problem occurs

[#11]

OK, so I haven't been able to get a Wireshark capture of what's going on, simply because the drops are too random and my main PC tends to stay wired (other devices don't have appropriate Wireshark apps). I've now hard wired both the WNR1000 and the DGN2200 to the DG834G to see if both exhibit the same problem at the same time. I've also updated the firmware on the DGN2200 to V1.0.0.36_7.0.36 and moved all routers out of their default IP ranges.
I'm now getting complete connection refusals on occasion, instead of a wireless connection but no connection to the router itself. Connectivity between the WAP and ADSL AP is fine as devices hard wired to the WAP can still connect fine. This is getting very annoying!
If problems persist then a change of ADSL AP is the next step.

[#12]

Well, after discussions with Matt Turner, our resident router expert, I think I've located and resolved the source of the problem.
The DGN2200 and WNR1000 both use a 40MHz channel bandwidth, which overlaps 6 channels, this is not compliant with IEEE 802.11n specifications. Other routers in the area don't recognise this signal and make no attempts to work with it, and happily blast their signal over the top, causing the problems that I've seen. Dropping the DGN2200 to 145Mbps changes it to 20MHz bandwidth and solved the problem. The WNR1000 would also need to be dropped down a notch - I believe it's listed as 70Mbps but haven't checked - in order to resolve the problem in the same way.
The reason for the two routers needing to be changed to different settings in order to solve the problem is because the DGN2200 uses 2x2 MIMO technology (read - 2 antennas) to double the throughput whereas the WNR1000 doesn't.