Hello Vishwa,
I would suggest to implement broadcast storm-control we use it and it help you to be able to access switches.
Actually, the right term is bridging loop: STP tries to provide a loop-free topology it doesn't create loops on its own.
Other features that can help:
bdpu guard on access ports using portfast
loop guard and root guard on inter-switch links
UDLD on fiber-based links.
When a real loop happen the signs are clear: main cpu usage rise to 80-100% and without any help devices become unresponsive to telnet.
Deploying the safety measures of STP and UDLD and the different forms of storm control (broadcast, multicast, unknown unicast ) help to give you time to shut down troubled link.
Storm control will discard traffic that will increment the output drops on the interface and can rise syslog message that can be seen on NME like HP Openview and OVO.
For example we limit to 1% broadcast traffic out GE ports.
Otherwise the only solution is to have someone unplug fibers or cable to break the loop and regain access to devices. And sometimes we had to do this way.
(a misconfiguration on a etherchannel caused the problem between two C6500)
Hope to help
Giuseppe
