L3 NSF on Cisco 3750 Stack

Unanswered Question
Feb 23rd, 2010

Hi there,  I have a stackwise pair of the 3750's running IP Services, one master and one member.  The stack is running OSPF and I have one OSPF L3 P2P on one switch and the same on the other, both equal cost.  If I fail the stack master by powering it off for example, I have to wait around 10 seconds before routing continues from the new master.  I have applied various combinations of NSF on the stack and the same on the opposite ends of the two L3 links, which actually makes it worse (20 seconds).  For further info, each stacked pair of 3750's connect to a pair of 6500 distribution switches.  Any assistance on the L3 NSF front is greatly appreciated.  I am not ruling out any bugs either in the IOS.

Thanks, Wayne

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Giuseppe Larosa Mon, 03/01/2010 - 12:27

Hello Wayne,

NSF use can lead to worse results in some cases.

With NSF you are asking to neighbor device to ignore the fact you have stopped to send OSPF hello messages and to hide the switchover.

It is probably better in your case to avoid to use NSF and use if possible small OSPF timers (sub-second hello messages)

I see NSF a good fit for a C6500 or C4500 with two supervisors installed on it.

To complete information you should provide the IOS image name running on the stack, the configuration of the stack and that of the neighbor devices.

Also the STP mode you are using on the client vlans, if in use can have an effect.

I guess you are measuring the convergence time from client vlan to remote IP subnet behind the C6500 devices.

Hope to help



This Discussion

Related Content