[Outages-discussion] Recent outage in Australia affecting Telstra

[Tom Storey]

Perhaps it would be a good time for everyone to remember the
importance of safe edge routing practices following an outage that hit
Telstra (Australia's largest telco) pretty hard a day or two ago. An
outage which could have very well been avoided, or at least minimised
in impact.

One of Telstra's downstream customers, a smaller ISP called Dodo,
accidentally announced the global table to Telstra (or perhaps a very
large portion of it.) Enough of it to cause major disruption.

Unfortunately it seems they had neither prefix filters, as-path
filters, or a max-prefix limit set on this customers BGP session.
Telstra's policy of preferring routes learned from customers
subsequently led them to try and reach the greater Internet through
said customer.

The general consensus is that having learnt these routes, Telstra
tried to send them onwards to its upstreams who did implement
max-prefixes and/or other protection mechanisms, and subsequently lost
most/all of their upstream capacity and disappeared from the face of
the Internet for a while.

So for everyone accepting routes from customers, make sure you have
your policies in order and dot all i's and cross all t's in your
configurations!

Im sure theres an applicable saying or quote about learning from the
mistakes of others that can apply quite aptly here, if this is not a
good enough example in and of itself how badly things can go wrong.
:-)