[nsp] Stateful Switchover

[Dan Armstrong]

Has anybody out there actually tested the SSO redundancy mode on a GSR?

During a maintenance window this morning, we decided it would be a good
idea to actually test it.  We pulled the active GRP-B out, and the
standby took over immediately.  Things worked for a few seconds, but for
some reason all of our BGP sessions reset, even though the standby was
forwarding traffic perfectly. We did have the bgp graceful-restart
command in place.

Weird.

Secondly, about an hour and a half after the switchover, the router
started "leaking" something.  IPs just started disappearing... slowly
over time.  Some IPs on interfaces were pingable and some were not.  It
seems that the stateful switchover did not work properly... Eventually
the router was not forwarding anything, and we had to reboot it again.

We have gone back to RPR+ just in case the active GRP-B does fail when
nobody is here to watch.  Yikes!

Any thoughts & experience are appreciated.

Dan.