HP ProLiant DL180 G6 2 NICs connected to Cisco2960S-Stack caused network fault after Windows2008

[Zer]

hello, I am a network engineer and not server admin. Just want to understand how it is possible.

Our server admin just installed a new HP ProLiant DL180 G6, connected 2 ethernet ports to Cisco2960S-Stack (Gi1/0/7 and Gi2/0/7 this is the ports on separate switches stacked in one logical switch Flexstack).

both switch ports configuration is:

interface GigabitEthernet1/0/7
switchport mode access
spanning-tree portfast
spanning-tree bpdufilter enable
spanning-tree bpduguard enable
end

After that he installed Windows2008R2 SP1. All was in default, interfaces was in default state (dhcp auto config).

One of the ports recieve ip address from dhcp server and will be available through the network, the other port did not recieve ip address (169.254...).

After that begins massive broadcast storm, that caused failure of the network, cpu utilization were near 95% at all switches. And I saw that switch port with one server interface had one mac address and another port had 2 mac addresses. After that i disconnects the server and all will be fine.

I'm not familiar with HP systems, any ideas, what can cause this storm, and how to mitigate this in future? 

[waaronb]

Without knowing more about how the server NICs were setup exactly, it sounds like a loop.

Try removing portfast from the IOS config on those ports. That's nice to have if you're 100% positive that a connection is just acting as a client, it's kind of risky, even if the switchport mode is set to "access".

Maybe whoever setup the server is trying to trunk the 2 ports using the network config utility, for instance.

I have 2960's for my servers, not stacked though. Each server connects to diverse switches.

When I was using Windows 2008, I used network config utility to do a simple active/passive NIC team since net throughput wasn't an issue. Now with Windows 2012 I use the built-in teaming and use Hyper-V mode for active/active (each of the physical boxes is a Hyper-V host).

But there are options in there for switch-assisted load balancing which could be causing trouble.

Although, it didn't sound like the NICs are in any kind of team since you said they're both setup for DHCP, whenever I see a switch do what you said, my first thought is always "there's a loop that STP didn't catch"