On 8/8/2012 4:14 PM, Steve Dalberg wrote: > CPU's were pegged for a customer of mine in California. tracked it > down to 2 events that went down at that time with a large message > volume. > > 1) Peering between GLBX and Level3 dopped somewhere, causing many > prefixes to shift away from L3 paths.
FYI: We saw significant route churn on GBLX (AS3549) as well as on Tata. > 2) Some IPv6 prefixes were aggressively bouncing from HE starting at > 11:25. Can't trace it any further upstream for HE unfortunately. > This cause lots of CPU churn on one of my customers networks > > May or may not be related. >