EMERGENCY MAINTENANCE OF ANL OC-12 AT 6:00 PM CDT

[Bill Nickless]

The Argonne router Kiwi is not properly doing "ATM ILMI registration" with
its ATM switch.  This is breaking the OC-12 network 192.5.171.0/28, which
at least one research group heavily depends on.  In particular, the host
named goshen.mcs.anl.gov at Argonne is unable to operate as part of the
Clipper testbed.  My attempts to fix this quickly last evening were
unsuccessful.

Goshen, up until about 5 minutes ago, had a default route via the OC-12
network 192.5.171.0/28 through an Argonne router called Kiwi.  This network
is broken, and has been broken for several days.  Until this default route
was changed, goshen was only available from hosts in the 140.221.8.0/22
LIS.  At this time goshen should be reachable from the Internet, but at
considerably limited data rates from its OC-12 potential.

I am planning to take Kiwi off line this evening at 6:00 PM CDT, or 4:00 PM
PDT, for diagnosis and repair of the 192.5.171.0/28 LIS.  This maintenance
window may last as long as three hours, depending on how events go.  The
ATM switch Mango may also be at fault, so it may be unavailable at times as
well.  During this maintenance window Argonne off-site traffic will go to
the secondary peering through the Argonne router Bluto.  THERE WILL BE NO
IP UNICAST ROUTING OUTAGE ASSOCIATED WITH THIS EMERGENCY MAINTENANCE.
However, please note that offsite multicast routing will not be available
during this outage.  I will be shutting down the external BGP peerings on
Kiwi throughout the testing process to avoid unnecessary flapping.

Please see http://www.anchor.anl.gov/doc/esnet.html for a drawing of how
the peerings are established.  During this kiwi/mango outage, Argonne
traffic will flow to ESNet through the peering labeled "Secondary Argonne
Peering".
===
Bill Nickless    http://www.mcs.anl.gov/people/nickless      +1 630 252 7390
PGP:0E 0F 16 80 C5 B1 69 52 E1 44 1A A5 0E 1B 74 F7     nickless at mcs.anl.gov