[c-nsp] Strange problem with Cat6500 freeze

[Robert Hass]

Hi

I have network where core-devices are Cisco Catalyst 6506-E with
Sup32/PFC3B. I last month We had two times problem. One time first 6500
'freezes' and second time second 6500 'freezes' Freezes means machine was
powered up, alarm was present (diode on supervisor), console wasn't
responding at all, freeze 6500 created a lot of loops on all VLANs inside
network (%SW_MATM-4-MACFLAP_NOTIF: Host xxxx.xxxx.xxxx in vlan XX is
flapping between port XX and port XX), all ports connected to 'freeze' 6500
was UP, LACP for PortChannels went down. Hard reboot (power off + power on)
helps both times. There wasn.t any crashdump in
flash/bootflash/sup-bootdisk/disk. Most problem was caused by loop created
by freeze 6500 - as all network was overloaded. How I could prevent these
issue in future ? Maybe storm-control for broadcasts ? Did anybody
occurred similar problem with 6500 ? After investigate both freezes was
probably caused by radom disabled/enabled NetFlow ('ip flow ingress') on few
SVIs.

Some facts more about both 6500 configurations:
- Both running 12.2(33)SXI4a IOS. Upgrade was done 1 month ago.
- Before machines was running 12.2(33)SXH4 never occurred similar problem
- eBGP (230k prefixes), 3-4 full table BGP peers + iBGP (20 peers rr-clients)
- IS-IS
- A little MLS QOS (policing) - 1-2 service-polcies on SVIs
- Control Plane Policing implemented
- Sometimes netflow v5 exporting from SVI
- Load around 40% (show catalyst6000) + 1.5Mpps (sh platf hardw capa pfc)
- Dual PSs
- Only ports GE on Sup-32 heavy used
- WS-X6408A-GBIC linecards but used only 1-2 ports with a little load
(~50-200Mb)
- ~300 VLANs
- ~50 SVIs

Robert