Hello,
If any of the switch fails, or faulty NIC, this brings the whole wireless network down, we have to pin point this client and shut the ethernet down, and when we go there there is either a faulty switch or NIC. But it totally brings the network on knees, nothing can be done, not even pinging, or at times we cannot even access the AP or SM to disable ethernet.
I am not sure if anybody else have experienced this problem, we might not be using enough filtering mechanism in our network. I am not sure what switch or what devices to use so that when this happens only certain parts are effected and not the whole network.
Any help ?
Many thanks.
Put routers at your towers
thanks for the reply.
the problem is we do bandwidth shapping from noc, and it woudl be hard to do it over router.
any suggestions what router models ?
if you want a high cost headached powercode.com has your solution, at this moment i dont recomend it
At some point we will need to implement routing at the towers and it’s highly likely we will use MikroTik to provide routing, packet shaping, etc. My plan for 2009 is to learn MT inside and out and them likely implement in '10 unless something forces my hand sooner.
thank you for the replies.
what is bugging me is i am not able to know what exactly creates problem, if just one client switch is faulty, they would bring the whole network down, i have one faulty switch with me, usually one of the port light stays on, regardless of connection, but when ever you plug that swtich on any other switch communication ceases, i tried to check with sniffers but no traffic flows.
What kind of switches are these and why are there switches at client locations?
Sounds to me like a flood of broadcast traffic.
If these switches are client-side and are the ones causing the trouble, why not just upgrade your SMs to the latest firmware (9.3) and put a limit of like 10k on broadcast traffic?
Sounds like broadcast problems. You will want to look into segregating your network as it sounds like the entire thing is a big bridge. You can accomplish this by VLANing the crap out of it or scattering routers about. I’m a VLAN & Just One Router kind of guy, but MikroTik is very popular in WISP-land. Basically a Single Board Computer with a custom Linux distribution. Another thing to do is throw the SMs into NAT mode. Turning each one into a router will eliminate broadcast issues from customers and standardize on-site configuration at the price of having to deal with port forwards and double-NAT issues.
Also curious about what kind of switches you are using. Are these D-Links from Best Buy?
This may be a dumb questions but, do you filter IPv4 multicast?
yes we filter IP4, Multicast.
8 port d-link switches, 1008. From D-link
It looks like this problem is with D-link and not everybody.
I tried to test, using a 24 port cisco switch, connecting 2 pc, pinging, once i connect this faulty d-link switch, everything is unreachable. no communications. even if i turn the faulty switch offf.
its not broadcast, i was thinking of broadcast, but it does not seem to be that.
You have a switch connected to the SM and then the PC’s connected to the switch?
Is the SM running NAT?
yes switch connected to SM, and then pc connected to switch.
and no NAT
possibly a bridging loop on the bad dlink switch? personally the best way to resolve the problem and keep it from happening again is use routers at the customers site or nat the sm. we use mikrotik routers at each tower site and require that the customer have a router in the home. this has saved us a ton of headaches dealing with broadcast storms and other issues such as people issuing dhcp across our network.
It’s bad practice to connect client machines directly to your network regardless if they have a public or private IP.
Either use a router or configure the SM for NAT. Then if something happens with the customer’s network behind the SM it will not affect your network as a whole.
what he said ^