On 6/8/11 11:35 AM, Mark Mason wrote:
This post would be for those of you in a DC environment with multiple internet feeds. When carrier X emails about a 12:00 - 6:00 maintenance window for IOS upgrade, reboot, card swap, etc. are you influencing your eBGP traffic away from that carrier during the maintenance window? Alternative being DO NOTHING and take the hard
In my past life as an NSP operator, I would always push traffic away from the affected link/carrier before the maintenance and restore default behaviors at sunrise or similar. Fearing uRPF and believing in stability where possible, I would change the inbound route-map and wait ~15 minutes (or until 'sh int POSx/y | i rate ' showed essentially no outbound traffic), then change the outbound route-map and call it a night. In the morning, I'd reverse the sequence but would typically shorten the delay towards four minutes or so. Management and customers were COMPLETELY oblivious to any carrier maintenance for years, including VPN activities.
Biggest tip is to write a set of route maps for maintenance "right away" and leave them on the router. In my case, we had two connections to each of three carriers, so I had -highmed route maps standing by for single-connection maintenances (-maint was for multi-site maintenances).
pt _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
