home 
wiki
wiki
Spelling: OSPFRevisited
This page exists as a depository for paraphrased conversation from the
mailing list, to allow a more permanent record of the discussion.
BACKGROUND INFORMATION overflow: auto; border: 1px solid; padding:
3px;"> Down in Hobart we've gone with BGP accross our entire network.
We initally tried OSPF about 2 years ago now but it caused major
headaches with regards to running stable. Quite a few times we had the
ospf daemons silently die but I think that was mostly due to multicast
issues.
At the moment we have around 9 or 10 AS on our network a map can be
seen here [14] We've been lucky in that our network has grown as a
single entity with no disconnected segments. Our assignment of AS is
fairly ad-hoc and no where near optiomal due to learning BGP while
setting it up but that's what it's all about in our books,
experimentation.
From a keeping the shit running point of view BGP makes much more
sence. Within a region or whatever you can run ospf/rip/eigrp (if
you're made of money etc but to get back into the main network you
need to talk BGP and redistribute your IGP into the EGP. It just so
happens that our entire network uses an EGP well iBGP
Config is fairly simple
router bgp 64512
redistribute connected
neighbor 192.168.4.9 remote-as 64516
neighbor 192.168.4.9 soft-reconfiguration inbound
neighbor 192.168.4.22 remote-as 64512
neighbor 192.168.4.22 next-hop-self
neighbor 192.168.4.22 soft-reconfiguration inbound
we set next-hop-self on iBGP sessions because we use /30 point to
point links. This came about due to the network starting off as an
ad-hoc card to card arangement we could move to adding subnets to aps
and assigning like that but it's never happend.
no need to set next-hop-self on eBGP links as that happens
automaticly when you cross an AS.
we don't run BGP on every node we only install it on a non endpoint
node. ie the site generally has an uplink and an omni and for end
users that don't have a second interface we add a static route and
redistribute static.
Quite a few of our routers these days are Linksys WRT54G's i think
we've gone from 0 to 9 of these units in around 2 months. we use
quagga on these units to do all of our routing and interface
configuration the only IP that is hard coded into the unit is the LAN
interface ip in NVRAM other than that they all come from zebra.conf or
via bgp.
QUAGGA LIST DISCUSSION
* Dan\'s initial message [15]
* James Haesu\'s response [16]
* Dan\'s response [17]
* James Hesu\'s response [18]
* David Young\'s (UCWN [19] Response)
LINKS
* BGP Introduction [20]
* CU Wireless protocol [21]
* WRT54G modification TasWireless [22]
Links:
------
[1] http://melbournewireless.org.au/?OSPF
[2] http://melbournewireless.org.au/?OSPF
[3] http://melbournewireless.org.au/?OSPF
[4] http://melbournewireless.org.au/?OSPF
[5] http://melbournewireless.org.au/?LocFinder
[6] http://melbournewireless.org.au/?BGP
[7] http://melbournewireless.org.au/?OSPF
[8] http://melbournewireless.org.au/?OSPF
[9] http://melbournewireless.org.au/?OSPF
[10] http://melbournewireless.org.au/?TonyLangdon
[11] http://melbournewireless.org.au/?DanFlett
[12] http://melbournewireless.org.au/?MarkAitken
[13] http://melbournewireless.org.au/?MelbWireless
[14] http://starnet.shacknet.nu/taswireless.jpg
[15]
http://lists.quagga.net/pipermail/quagga-users/2004-December/003350.html
[16]
http://lists.quagga.net/pipermail/quagga-users/2004-December/003351.html
[17]
http://lists.quagga.net/pipermail/quagga-users/2004-December/003362.html
[18]
http://lists.quagga.net/pipermail/quagga-users/2004-December/003365.html
[19]
http://lists.quagga.net/pipermail/quagga-users/2004-December/003375.html
[20] http://www.cisco.com/univercd/cc/td/doc/cisintwk/ito_doc/bgp.htm
[21] http://www.cuwireless.net/downloads/ETXprotocol.txt
[22]
http://starnet.shacknet.nu/wiki/index.php?page=LinkSysModification
[EditText] [Spelling] [Current] [Raw] [Code] [Diff] [Subscribe] [VersionHistory] [Revert] [Delete] [RecentChanges]