Hello,

We tried to build a mesh system with 7 AP's;

First we used the simple approach with WDS and rtsp - it worked, but the performance was bad.

Then we tried using HWMP+ and it seemed to perform much better. But after a few minutes the latency to all devices goes up and up. A few seconds later none of the devices responds anymore. Even the one i'm directly connected with per ethernet does not respond.

Can someone point me in the right direction - whats wrong? I tried nearly every option, but the problem stays.

I used firmware 3.26 and 4.0beta3 - same behavior.

Thanks

We have been testing HWMP+ on each and every release hoping that this problem would be resolved.

The issue appears to be in the FDB updates. FDB is how paths across the mesh are determined. Basically a spanning tree like update keeps a forwarding table current for each node. But at some point the FDB stops getting updated. Paths are lost, and the system collapses.

The protocol is way to erratic to be used.

We found that if you just use dynamic WDS with RSTP you get the same or better results.

We have tried overlaying OSPF, using EoIP tunnels, and all other manner of schemes to get this to work without any success and have given up on HWMP as completely unusable.

We would love to see this fixed too. It is the holy grail of wireless for MT. I don't think it will ever work properly.

Thank you;

It's good to see i'm not alone; and that i've nothing misconfigured.

Hope it will be fixed soon...; i need it up in some way before september.

BTW.: It seems WINBOX interferes with HWMP+ somehow. Whenever i connect to any of the devices with Winbox
the network crashes. I can reproduce this behavior anytime i want - it alway crashes...

I am guessing that you are trying to use MAC address connections when Winbox crashes. You are probably initiating a FDB update by trying to connect via MAC. This causes a network flood which causes everything to crash.

First, HWMP is not ready for prime time.

But even so, you should be able to connect using IP address. Don't open the ... connect list, just type in IP address and go. It should work.

It looks like you have a routing loop. It could be bug in the implementation, but also it could be also if you put together meshed network with a bridged network. The routing loop could happen if the mesh network is connected to a (R)STP network in more than one place.

Yes, i think thats exactly what happens. But i can't imagine how that loop is produced.
I only have wireless devices connected at the moment - the 7APs and one wireless client.
No additional bridges with (r)stp in place.

@DogHead: Yes you're right. It works perfectly as long as i connect per IP not per MAC. Thanks again.

So as outlined in several dozen previous messages, we have tried every conceivable configuration to get past issues with HWMP.

All configurations we have tried to rely on dynamic mesh wds connections between mesh nodes. We have tried putting WDS interfaces into a mesh interface, putting them into bridge and then into mesh, not putting the physical wireless interface into the mesh or putting it in. We have tried routing OSPF over this, we have tried using EoIP tunnels, we have tried bridge filters, static ARP...in other words we have given it a good shot.

The problem is that HWMP just does not work correctly in keeping the FDB updated, which is the whole point. There seems to be flooding that kills the network and then takes a long time to build back up.

We have sent supouts on this, documented our network configurations.

But we have given up.

Tell us when it is really working. Set up a star mesh of 6 nodes with clients hanging off the ethernet of each and then pass traffic between all of the clients over the mesh. If that works, we will be very happy.

Hi There,

Using now 3.28 - HWMP+ in a live network with dynamic WDS. I am using Crossroad boards with only one antenna for dynamic WDS links between them and VAP on top of those for Client Access. All on same frequency. WDS is using different SSID and WPA encryption as virtual access point. Some interlinks are done with RB 433.

There are currently 12 devices running and most of those are seeing at least 2 WDS connections. After crashing the system now for the past 2 month, i finally got it stable. i am using only one portal which is running on a RB450G and hotspot on it. i have no chance to try 2 or more portals at this stage.

one issue i would like to mention: I did not understand for a long time the difference between pro-active and re-active HWMP+ setups! As most of my traffic is internet based, i choose pro-active mode!

my mistake was to think i can run BOTH setups together, By default MT's are configured to workin in re-active mode. As soon as I enabled a portal, both modes were running and eventually crashed the system as described earlier on

the solution was as following (running with 1 portal)

hwmp-preq-destination-only=NO (default yes)
hwmp-preq-reply-and-forward=NO (default yes)

My understanding is that both this values are only needed while running re-active mode without having a portal configured.

hwmp-prep-lifetime=30s (default 5m)

Right now, there are 55 users online, no confused mesh FDB entries, running since 6 days. i am happy and getting some sleep!

M

what did you ment with 'crashing'?

Those changes that you made below, decreased the mesh protocol traffic to minimum. Probably in future we will use those values as default.
hwmp-preq-destination-only=NO (default yes)
hwmp-preq-reply-and-forward=NO (default yes)

Your effort encourages me to try HWMP+ aswell..
Thanks swissiws!