brycec: so if you were behind one of those 36 other networks, that peer with my network, then you could still get traffic through.  the problem is it would be unlikely you would be behind those networks since those peers are mainly other data center ISPs, not residential / business ISPs (e.g. time warner, comcast, cox, etc..., to name a few US examples)
JP seems to have good IPv6.  Many orders from Japan are now exclusively from IPv6 addresses.  Very interesting.
inet6num:       2001:0DC2::/32
netname:        JPNIC-NET-JP-20030529
descr:          Japan Network Information Center
up_the_irons: I see... And in the case of yesterday's outage, being behind HE who peers with you but had problems...
I'm once again a little unclear on why the failover didn't work? Was it HE that didn't find a different route to reach you? Was it your router for not finding a working route to return my packets?
brycec: sounds like (if I understand properly) he.net will not reach arpnetworks in the event of an arpnetworks <-> he.net outage, seems counterintuitive to the standard mentra of 'route around the issues' and how I'd expect bgp to work. maybe someone more clueful than me can explain more details on why ;-(
toddf: exactly my initial interpretation, but up_the_irons said otherwise. 20:30:17 >>@up_the_irons<< brycec: No, if the HE node is down, traffic would get re-route elsewhere by BGP.
something was down other than the 'HE node' I guess, not sure what exactly that all means either. tis why up_the_irons runs arpnetworks and I can't hope to do the same ;-)
heh, I think I've got a decent grasp on the concepts and implementation of virtualization, basic networking, even ipv6. But BGP is right on the edge of my understanding (I get the concepts, but not the ways it might fail), and peering/transit are just plain magic to me at this point. Bandwidth happens, and BGP tells the world how to reach us over which pipes...
and yet, somehow, it failed to re-route around whatever arpnetworks <-> he.net issues arose
So from my simplistic point of view - One provider went down, another was found to be down, but we have others and BGP should have magically made everything reroute and work.
heh can't wait till second year networking courses, learning BGP :)
thats about as far as I grasp
I feel like the blind is leading the blind in this case ;-(
*shrug* At this point, no use in hounding up_the_irons with more questions. He's getting it fixed and won't let it happen again. RIGHT?
good book on the topic - http://shop.oreilly.com/product/9780596002541.do
I grasp enough to know he's looking at making it less likely but I'm not fuzzy with the whole he.net failed to let us reach arpnetworks scenario
there was no transit path *back*, wrt IPv6, b/c HE and Trit were both not routing our traffic.  as a result, all IPv6 was unreachable even though other peers were up (unless you're  behind those peers, which is unlikely)
HE and Trit give me a full routing table