Date: Tue, 14 Apr 2009 18:56:08 -0400 From: Steve Bertrand <steve@ibctech.ca> To: Charles Sprickman <spork@bway.net> Cc: freebsd-isp@freebsd.org, "Justin G." <justin@sigsegv.ca> Subject: Re: BGP with OpenBGPd. Message-ID: <49E51488.9010202@ibctech.ca> In-Reply-To: <alpine.OSX.2.00.0904141652300.579@hotlap.local> References: <5da021490904131135k7c78b2few5c48ee8b0a001e5@mail.gmail.com> <alpine.OSX.2.00.0904131555580.49636@freemac.nat.fasttrackmonkey.com> <49E489EB.2090802@ibctech.ca> <alpine.OSX.2.00.0904141652300.579@hotlap.local>
next in thread | previous in thread | raw e-mail | index | archive | help
Charles Sprickman wrote: > On Tue, 14 Apr 2009, Steve Bertrand wrote: >> I don't understand how they "sound" buggy. What exactly are you >> referring to? Which pieces are you concerned with? > > I'm probably reading too much pro-OpenBSD stuff. :) On the > OpenBGPd/OSPFd pages there are a good number of technical presentations > where they explain how their design diverged from the existing open > source routing daemons. > > I also occasionally peruse some WISP forums, and have seen some horror > stories in there, since those guys rely very heavily on homebrew hardware. I've heard things as well regarding the back-and-forth of Quagga vs. OpenBSD, but as I said...Quagga works, for *my* environment without any issues whatsoever. >> I've yet to find a bug. Every time I think I've found something, it has >> come down to a simple inconsistency between how I'd do the same thing on >> a Cisco IOS. > > That's understandable, and something I'd also have to deal with on a > used Juniper. FWIW, I can grab loaded M20s for about $6K each. It's > very hard to say no at that price. Agreed. However, considering you are already contemplating using PC hardware for the job, I'd still lab-it-up. There's not much better feeling than knowing that you have backup hardware at your fingertips if a router fails. Even at $6k a pop, the budget police usually don't understand line-items such as 'spare routers' :) >> We run ~1/2 of our routers on FBSD based hardware that run from either >> USB thumb stick, or CF/SD cards. > > Stock FreeBSD or do you pare it down? Years ago, it was quite pared down, to < 32MB. Now, it's generally a stock install with a modified make.conf that trims documentation and other cruft. > How do you handle upgrades? > Install on another flash card and just reboot to the new card? Depends on the upgrade. They all begin the same way, by inserting another flash card and using "dd" to image the running system onto the backup. I do this twice onto two separate cards. In my lab, I have a few PC's that I use exclusively for testing routing functions. Some are build boxes, and others are actually used for booting the routing system. For upgrading routing daemons, I insert the backup card into one of my build machines (that runs it's own host OS), mount the card to /mnt, and generally install from ports (using the hacked make.conf). Then I remove the upgraded flash card, boot one of the lab routers with it, and test to ensure all is well. I then (on the running test router), insert the secondary backup card, and copy ONLY the modified files (usually only binaries) on top of the old files manually. I reboot using the secondary card as the boot device, and if things come up, then I simply copy the binaries onto the production router, and reload the daemons. Upgrading the daemons does not happen often, but when new functionality comes out (such as md5 for bgpd), this system works fantastically. For the OS, a similar approach is used, but since I do have to reboot the router for kernel changes, I do simply swap out the memory card to a tested, upgraded one after building it on the build machine with DESTDIR=/mnt. Doing the OS doesn't happen all that often either though. I treat the FBSD routers the same way I do the Cisco's, using management & control plane access restrictions everywhere. The FBSD routers are simply that...routers. I've been typing pretty quickly while thinking about other things, so I'm sure that there are pieces I've missed ;) >> As for your OC3's: >> >> http://www.prosum.net/atm155_E.html > > Wow. Those list for what looks like under $1K US. Impressive. Our DSL > provider is actually going to be moving from giving us an OC-3 for > customer backhaul to a GigE handoff. Details of how this works are > still murky though - if they're going to do a VLAN for each customer, > I'd think they'd run out of VLANs before running out of bandwidth... We don't do DSL out of our PoPs, as we pretty much wholesale through a couple of other companies. That said, I know two of which aggregate over Gi, so I'll ask them for info regarding how the separation works. Cheers! Steve
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?49E51488.9010202>