IRC Logs

Network: freenode Channel: #cisco

	Enter your search terms Submit search form
	Web www.irclog.org

[00:58]<sazdn>this keyboard is a POS
[00:58]<2zrgvj-l2>nemith, isn't the use of vpls just a software thing?
[00:59]<msxjv>obligatory walk to Starbucks, werd
[01:00]<msxjv>nosx, you got your stuff squared away it looks like?
[01:01]<gndyvx>branto-pb: i thought vpls required some real hardware
[01:01]<gjsw>lol dshot, this is so random. i leave it for 5 minutes and it fixes itself on 1 ibgp session.
[01:02]<2zrgvj-l2>not sure... I thought you might be able to get away with doing it in sw... could definitely be wrong
[01:02]<gndyvx>or even atom
[01:02]<sazdn>i thought vpls was a methodolgy, not somthing tangible
[01:04]<gjsw>it just randomly spit out: "04:41:34: %BGP-5-ADJCHANGE: neighbor 3.0.2.1 Up"
[01:04]<gjsw>whatever was broken seems to have fixed itself? idk lol
[01:05]<msxjv>heh
[01:05]<msxjv>sounds like you had some synchronization issues
[01:07]<msxjv>basically, if you don't have an IGP route in the routing table that matches what you want to propagate via IBGP it causes woes
[01:07]<msxjv>did you do a 'sh ip bgp x.x.x.x'?
[01:07]<gjsw>yes, and thats the interesting thing.
[01:07]<gjsw>eventhough each ibgp peer had a working connection to their external peers, and eachother
[01:07]<gjsw>they didnt seem happy
[01:08]<gjsw>until i provided them with an OSPF path to the OTHERS external peer
[01:08]<msxjv>aye
[01:08]<msxjv>if you look at the route, it should say something about not being synchronized
[01:08]<gjsw>it said it wasnt being distributed, which was the funny thing.
[01:08]<msxjv>which is a tip off
[01:08]<gndyvx>redistributing between igp and egp is a bad pratice these days
[01:08]<msxjv>yeah
[01:08]<gndyvx>but if you are not doing next-hop self, you will need the external peer in the igp routing table
[01:08]<gndyvx>yes
[01:08]<msxjv>sync off is always pretty standard
[01:09]<gjsw>im running no sync right now as directed above.
[01:09]<msxjv>unless you like having 200k routes in your IGP
[01:09]<msxjv>lol
[01:09]<msxjv>=P
[01:09]<gjsw>im redistributing bgp into the OSPF and EIGRP routes for the internal network.
[01:09]<gjsw>=/
[01:10]<gndyvx>if i am not doing next-hop self between ibgp peers then the next hop is not modified in the prefix so the further end router has to do a route lookup to find where to send the packet
[01:10]<gndyvx>if it can't find it in it's routing table it will drop it
[01:10]<gndyvx>i like redistributing only a default route into my igp
[01:10]<gndyvx>makes for a resilient igp exit point :)
[01:11]<gjsw>right now im just testing with a pile of spare 2500s, so im not too worried about routing table size, however i dont think my current solution would work if i were to try and peer in at an IBX =P
[01:11]<gjsw>i would forsee alot of static routes to fix the discontinuities.
[01:12]<gndyvx>which is bad as well
[01:12]<msxjv>hmm, you mean redistributing statics via IGP?
[01:12]<gjsw>just statics to make ibgp happy =P
[01:13]<gndyvx>you should run ospf or is-is as your igp and be careful with it
[01:13]<gndyvx>usually you just want connectivity for the routers and nothing beyond
[01:13]<sazdn>i prefer ospf
[01:13]<sazdn>but ymmv
[01:13]<gjsw>i simplified the setup i posted earlier, in reality every AS has 2-3 routers in it and there are 4 AS's. im running EIGRP in one transit AS, and OSPF in the other transit AS.
[01:14]<gjsw>I havent actually played with IS-IS before.
[01:14]<gndyvx>is-is is used by a lot of big isps
[01:14]<msxjv>also nosx, if you're testing and you're leaving in synchronization, be sure the RIDs are identical between BGP/OSPF
[01:14]<sazdn>i avoid EIGRP because its propreitary
[01:14]<gjsw>synch is off @ all routers right now dshot.
[01:15]<msxjv>yeah, as much as I like the idea of EIGRP.. I usually don't use it much
[01:15]<gjsw>EIGRP is fine when your in a hurry and need a simple solution that doesnt involve rip =P
[01:15]<msxjv>here and there in times past
[01:15]<sazdn>ospf is just as fast to whip up
[01:15]<gndyvx>eigrp acutally has some limitations
[01:15]<gjsw>OSPF requires 2% more actual thought to impliment =P
[01:15]<gndyvx>beyond it's propriotaryness
[01:15]<msxjv>eigrp gets nasty with larger scenarios
[01:16]<msxjv>query process, lack of hierarchy
[01:16]<gndyvx>SIA's
[01:16]<gjsw>my understanding was isis is still the same sort of link state shortest path algorithm, whats the big difference between it and ospf?
[01:16]<msxjv>yeh
[01:16]<gndyvx>nosx: not much
[01:17]<gjsw>does it perform better in multi-area situations or something that would appeal to larger network operators?
[01:17]<msxjv>nosx, you should check out some of the NANOG presentations
[01:17]<msxjv>AOL had a good one for the switch to IS-IS from OSPF
[01:17]<gndyvx>http://catfive.chaosnet.org/why-is-is.html <-- see our own Catfive's writeup on it
[01:17]<gjsw>Ill queue the videos for download later
[01:18]<gjsw>nanogs video archive is really a handy thing to have.
[01:18]<sazdn>isis was developed without IP in mind, OSPF with IP in mind
[01:18]<gndyvx>true
[01:18]<sazdn>ISIS doenst use IP to communiate
[01:18]<msxjv>I'm not versed terribly well in IS-IS and no real operational experience with it personally, so really my own comments aren't of value
[01:18]<gndyvx>isis was designed to be extendable
[01:18]<gndyvx>ospf was not
[01:19]<sazdn>which is why it went to IPV6 a lot quicker and easier
[01:19]<sazdn>and OSPF needed a re-haul to OSPFv3
[01:19]<gndyvx>exactly
[01:19]<gjsw>Speaking of V6, whats the current plan for AS space? Alot more people are going to want to be multihomed in the future, so are they moving to a 32bit AS space?
[01:20]<gndyvx>yes
[01:20]<gndyvx>there was an article the in IPJ about it
[01:20]<gjsw>Seems the last time i asked the plan was changing every other weekend.
[01:20]<msxjv>that's how wars are fought
[01:20]<sazdn>they need to go beyond the 65k ASn's availble anyway
[01:20]<gndyvx>http://www.cisco.com/web/about/ac123/ac147/archived_issues/ipj_9-1/autonomous_system_numbers.html
[01:21]<gndyvx>it will be fun to read AS Paths with dots in them
[01:21]<gjsw>lol
[01:21]<gjsw>It will be fun to try and read somebodys ip in v6 format =P
[01:22]<sazdn>well 64509 ASNs
[01:22]<sazdn>ipv6 wont be as hard to implement as people think
[01:22]<gndyvx>nah
[01:22]<sazdn>most ISP's will only ever get 1 block assigned to them
[01:22]<sazdn>so thats a $PREFIX
[01:22]<sazdn>the last 3-4 ocets go to the customer
[01:23]<sazdn>so you only gotta worry about 2-3 blocks to remember
[01:23]<sazdn>sextets?

Back to #cisco
Or go to some related logs:
#debian
#blender
#gentoo
#gentoo
#ubuntu