08-24-2017 03:08 AM - edited 08-24-2017 05:43 AM
We have some BR6910 switches here but we are facing some cpu usage issues when we create some connections between them.
I've attached an overview of this connections and when we enable ports SW03/P11 or SW02/P11, the rstp sets the ports SW02/P11/P12 as alternative discarding but the cpu usage on switches increase from 10 to almost 90% (even 100%).
I can't see what process is causing this cpu usage because this switch doesn't show this, it show only the total cpu usage (show process cpu).
Has anybody faced any issue like this or does anybody have any idea about what is the problem?
08-30-2017 02:26 AM
This does sound like you have a loop created when you enable the resilient links.
You could probably verify if a loop is causing the high CPU by looking at the interface counters to see if you are seeing a high rate of broadcast or multicast packets which would be indicative of a loop.
Have you verified that all VLANs are configured correctly for spanning tree, all VLANs will need to be under STP control?
08-30-2017 04:41 AM
Thanks for replying!
I've checked the rstp configuration and (at first time) all vlans are procted. (images attached)
I've noticed a lot of multicast and some broadcast packets on interfaces, but I didn't understand why if rstp has blocked the interface on sw03.
Could you help me?
08-30-2017 05:14 AM
I would suggest that you verify that counters are not incrementing whilst ports are disabled and that once ports are enabled and CPU is high verify that the counters for multicast and broadcast are incrementing, may be worth clearing the counters each time to verify by how much they are incrementing.
Then I would verify that all interfaces are under STP control and all loops are blocked, in your topology there are 3 potential loops 2 x triangles one outer square (hope this makes sense)
Its simply a matter of troubleshooting one step at a time.
if everything looks OK after all these checks I would suggest that you open a support case with Brocade to troubleshoot problems further
08-30-2017 10:40 AM - edited 08-30-2017 01:13 PM
All counters were cleaned before I've taken the "print screens", so these information are related to 15 seconds wich I've left the interfaces up.
Spanning-tree is enabled for all interfaces and is blocking correctly the interfaces when we put the interfaces up.
Unfortunately we don't have any active support contract and I think we can't open a support case with Brocade.
Thank's for your time!