[c-nsp] Reaction to forwarding failure...
rodunn at cisco.com
Thu Oct 7 10:55:16 EDT 2004
We are going to create a new track object that the
higher layer protocols can hook on to as the singal.
That way HSRP could failover and the routing
protocols could listen to the notification and just
make the interfaces passive to the routing protocol.
Externally found enhancement defect: New (N)
Request for a new tracking object to monitor for CEF disable event
While I haven't tried this yet I think you could
do the same thing already with EEM and watch the syslog for
the FIBDISABLE message and then just either shut down
the BGP peers or make the interfaces for the IGP passive.
EEM with the TCL scripting has the capability to be flexible
for a lot of scenarios like this as well as troubleshooting.
It's pretty new.
On Wed, Sep 15, 2004 at 06:08:19PM +0000, rwcrowe at comcast.net wrote:
> I for one would love this feature. I see alot of designs where L3A and L3B are setup as you have shown or they are dual-homed to CoreA and CoreB. But as you said if a PFC or some forwarding engine fails you are stuck with process switching and your performance goes into the toilet.
> rwcrowe at comcast.net
> -------------- Original message --------------
> > I had an idea and wondered how common this would
> > be and how much customers would like it.
> > The idea is how to react on a known failure
> > type for a redundant design.
> > Let's say you have this:
> > COREA COREB
> > | |
> > L3A L3B
> > | | (HSRP)
> > Access layer switches
> > Pretty typical design if you are not doing L3 all the way to
> > the access. Now let's say you have some form of hardware
> > forwarding failure on the L3A (l3 switch) switch. With the
> > failure there is a chance you punt the packet to process
> > level and overrun the CPU. Assume L3A is HSRP primary.
> > What about if there were a configurable option that for
> > known failure conditions you could have all routing
> > disabled for the routing protocols and also have HSRP
> > disable itself? That way you would failover to the
> > redundant path both ingress/egress to the core.
> > Clearly this doesn't apply to all designs. If you only
> > have a single path you would want that path to continue
> > to pass the traffic that it can. You could still reach
> > L3A from COREA by telnetting to the directly connected
> > ip address.
> > Thoughts?
> > Rodney
> > _______________________________________________
> > cisco-nsp mailing list cisco-nsp at puck.nether.net
> > https://puck.nether.net/mailman/listinfo/cisco-nsp
> > archive at http://puck.nether.net/pipermail/cisco-nsp/
More information about the cisco-nsp