On Tue, 27 Jun 2006, Jeremy Chadwick wrote:
Was anyone able to get an RFO or post-mortem for this?
"An inaccurate set of BGP policies were distributed to routers connected to AS791 1 when an automated update script ran at 1100 MDT. The update script regularly ran every two hours to update the network with current BGP information. Due to the scheduled shutdown of the legacy BGP policy server and subsequent con version to the Level3 route registry engine, the old server policy server was sh utdown. In addition, the scripts used to update routes on the network were to be disabled. One of these scripts wasn t disabled as intended. As a result, the script ran as scheduled at 1300MDT and consequently pushed partial configurations to production routers because the script was unable to communicate with decommissioned policy server. Incorrect policies were exchanged between AS7911 s customers and peers resulted in increased latency; as large route blocks attem pted to traverse individual customer connections. Repair Updated configurations were pushed to all the routers, individual connections were cleaned up and BGP sessions were restored. In addition, the automated BGP script has been shut-off. Maximum pre-fix list limits have been established across the network as a risk mitigation step. " --- david raistrick http://www.netmeister.org/news/learn2quote.html drais@atlasta.net http://www.expita.com/nomime.html