06-03-2014 09:47 AM
We have a pretty simple network, simple hub and spoke with a RX-4 at the center and several layer-2 switches, one to each port on the RX-4.
My question is around storm prevention. We had a student plug a wire into the wrong port on a switch and created a storm that took quite a while to solve (I ended up pulling wires on the RX-4 track down the problem)
I would like to be a lot more proactive both in preventing a reoccurance and in my ability to find the root cause. There are a lot of configuration options in the Brocade manual around limiting broadcasts however they seem to assume spanning-tree and we do not have that enabled. so my questions:
1) what options should I have enabled at the RX-4 and my layer-2 switches for storm prevention when spanning-tree is not enabled?
2) The manuals state that you can enter a value for things like "broadcast limit" but how would I know a proper value to enter? is there a baseline or way of tracking "normal" numbers somewhere?
3) finally there is mention of sending a trap when it happens but no mention of what trap to listen for, is that documented anywhere?
Any help would be greatly appreciated!
06-21-2014 05:02 AM
You need to create a baseline for your network, sFlow would be a good place to start. However you can use 'show statistics ethernet slot/port-no' to see what is being collected on the RX - you should clear the stats first so you know how many you are getting while you do the baseline.
Any reason not to use some form of STP? You can still rate limit on RX e.g.
BigIron RX (config)# broadcast rate-limit 1000000 1 np 3/2
Also even if the RX is hit by a storm you will still be ble to access the device via the management port (it will be slower but will still work). Is this way when you do a 'show cpu' you will see you are being hit by a storm.