I am recieving this "Kernel: Network packet flooding detected." error a few times per day. This is being reported on my c7000 in the log. When this occurs all of my blade servers (bl460 g1) peg 100% utilization for about 30 seconds, then return to normal. Trying to resolve this I updated my onboard admin to 2.6 from 2.52. The iLo versions on the blades are all 1.78.
This means that there is excessive traffic being seen on the OA/ILO network port that connects to the rest of your network.
Do you have a separate management VLAN for the OA/ILOs? That would be a good idea to isolate the OA/ILO's from your normal production traffic.
Otherwise check for excessive broadcast/multicast traffic or unicast flooding on your network. Excessive Unicast Flooding can be caused by spanning tree topology changes as everytime a spanning tree topology change occurs the mac address tables of your switches are flushed and everything is flooded until it is relearned.
For multicast try to use IGMP snooping.
take a network trace to identify the nature of your excessive traffic.
Thanks for the info! I have found that it does appear to be related to traffic on the core switch, I have not yet setup a separate vlan for the iLo management, but plan on doing this soon.
Currently spanning tree is off on this switch. Would using a product such as wire shark to trace traffic be sufficient?
FYI everybody - This issue has not been fixed by HP. It has happened in our environment in v2.21, v2.25, v2.41, v2.32, v2.51, and v2.52 and continues to happen. I believe is a hardware issue. Make note that HP considers these OA's obsolete now. They are now pushing the Gigbyte version now.
I am having same issue; HP advised to have OA port on seperate VLAN than rest of Blades, HP advised that the OA is getting hit "hard" but not seeing the traffic on the cisco switch side nor blade side.
This message gets logged to the OA System Log when a certain packet per second threshold is detected on the OA NIC. I'm not certain but I've heard that the threshold is something like 10000 packets over a two-second period.
This can be verified with a protocol analyzer listening to the OA/ILO port.
I looked around in the OA 2.60 GUI and CLI to see if there was a threshold that could be adjusted but couldn't find anything...
