Hi All,
Forgive me for the long winded email but I have a situation that I
would like to hear yours suggestions / recommendations on. Here is the
following scenario:
We have users that have a frame-relay PVC that terminate on a
CBX-500. That CBX has configured what is called a "redirect-pvc" which is
essentially one main PVC that you specify a primary and backup PVC for. The
primary PVC goes to an M20 over OC-3 and the backup PVC goes to another M20.
There is a logical PVC built for each user from the CBX that the user
frame-relay terminates on to each of the M20's. Each of the M20's are
configured identical so that in the event that the primary PVC goes down, it
will redirect it's traffic using the backup PVC to the backup M20. We were
using the OAM-Period command on the M20 so that it can notify us when the
PVC goes down. The oam-period command works by sending F5 loopback cells
end-to-end however, since the end CPE does not respond to the OAM F5 cells
the Juniper thinks the PVC is down and it will take its logical interface
down. If we turn the oam-period command off then the Juniper does not
depend on the F5 cell for status of the pvc but instead it will rely on the
AIS cells to tell it when the pvc is down. The problem comes in place in
the event that the entire redirect-pvc gets deleted from the CBX-500. In
that case none of the M20's will have a way to be notify that is down
because they do can not receive AIS cells. This causes both the M20's to
think that the pvc is up (since it does not receive notification stating
otherwise) and it can cause blackhole routes for some customers network who
are being advertise with aggregate.
Juniper has the Ping ATM segment command that will send F5 OAM
loopback cells to the immediate ATM switch. I would think that if we can
get that same functionality available on the logical interface to be able to
change the default end-to-end behavior it will notify the Juniper that the
PVC is down in the event that it is delete because it will know that it is
not receiving the F5 Loopback cells back.
Perhaps my logic behind the M20 bringing up the interfaces is
incorrect but please advice if otherwise or with suggestions.
Thanks in advance.
Mario Puras
JNCIS, CCNA, LSCP, MCP
SoluNet Technical Support
Support Engineer
This archive was generated by hypermail 2b29 : Mon Aug 05 2002 - 10:42:37 EDT