[Gllug] Routing problems

[Mark Lowes]

On Thu, 2002-06-27 at 18:59, Adrian McMenamin wrote:
> Jun 27, 15:48
> malcolm
> The current routeing problems have been traced to an apparent failure of a 
> switch at the LINX. Most UK ISPs will be affected along with external 
> Internet connectivity.
> 
> ===============
> Anybody else having problems? Surely not just one switch? Is there so little 
> redundancy built into the internet in the UK?

It's a major problem, the faulty switch is a core one (handling the
inter-vendor link), so anyone on the affect switch gets a session flap
(100+ sessions bouncing is not a fun thing to happen to any bgp router),
due to the construction of any switch network if you lose a core switch
in a ring the ring has to either (a) renegotiate STP or (b) humans have
to live up the link in admin-down.  The LINX is physcially built up in
two rings (one foundry, the other extreme).

Note that the LINX is heading for the 20Gbs level[1] and is therefore
breaking some interesting ground (they've recently rolled out 10Gbps
interswitch links.

On the plus side the LINX (and all members) has been undergoing a
reconfiguration over the last two weeks which is putting in place some
changes which will limit the impact of a crash on one of the rings. 
Unfortunately this failure in the switch happened about a week before
the planned removal of the foundry<->extreme link.  Sods law really.

   Mark

[1] https://stats.linx.net/

-- 
The Flying Hamster <hamster at korenwolf.net>     
http://www.korenwolf.net/
"It was our first attempt; we didn't get nearly as far as astray."
   -- Onslow, "Keeping Up Appearances"
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 192 bytes
Desc: This is a digitally signed message part
URL: <http://mailman.lug.org.uk/pipermail/gllug/attachments/20020627/d70287cf/attachment.pgp>