Friday was fun. A machine was spoofing and attacking another machine, and it took out a couple routers and the internet link. After that was done there were still problems. We ended up finally tracking down the problem with high latency lately. Suffice it to say the HP Procurve 4000s suck for certain tasks. Like mirroring ports under any load. It was losing about 40% of the traffic on a mirrored port. It wasn't noticed, because it was clamped on both sides to about 9Mbps (physical link on one side, packetshaper on other). When sending through floods that the packetshaper was instructed to ignore, it'd be 35-40% loss. Other traffic 0% loss, but a couple hundred ms longer rtt than there should be. Apparently the issue was being handled at layer 2 so nothing on either side ever saw it. Even VoIP traffic worked great - it's amazing how much delay people have become accustomed to on the phone. Anyways, after turning off the packetshaper (last device before the packet loss), physically bypassing it, replacing the ethernet cable, and getting ready to swap out that vlan's ports on the switch with a cheap one to see if it was acting wrong (last step since the traffic already went through that switch in another vlan with no problems), we tried turning off the mirroring. After further analysis of the port counters, the port it was mirrored to had over 10mil tx errors in the time the switch had been up (recently rebooted due to the other issues). Fun, huh? At least it's fixed now.