Re: Multipath to CISCO
On Wed, Nov 5, 2008 at 2:11 PM, Jussi Peltola [EMAIL PROTECTED] wrote: The other option I believe would be using PF to round robin the packets on both destinations using route-to rules. Would this work? Why wouldn't it? Not that I can think of, I guess that is why I am emailing the list :) Just more wondering if anyone has had experience on this specifically. I'm really trying to avoid having to buy an 1841. Don't have the budget or inclination to spend that much money for a little green box when I think my OpenBSD box can handle it. If you have two ethernets and you want to round-robin, trunk(4) might work too. Yeah, that would work I guess, if I could get the modems into bridge mode and then talk directly to the central cisco. not a bad idea. -- http://lindsaar.net/ Rails, RSpec and Life blog
Re: Multipath to CISCO
On 2008-11-05, Mikel Lindsaar [EMAIL PROTECTED] wrote: The other option I believe would be using PF to round robin the packets on both destinations using route-to rules. Would this work? it should, but you might need to make the rules stateless (no state).
Re: Multipath to CISCO
On Wed, Nov 05, 2008 at 09:40:02AM +, Stuart Henderson wrote: On 2008-11-05, Mikel Lindsaar [EMAIL PROTECTED] wrote: The other option I believe would be using PF to round robin the packets on both destinations using route-to rules. Would this work? it should, but you might need to make the rules stateless (no state). It works, and you do. # san2 and san3 are in interface group att att_if0=san2 att_if1=san3 pass in log on att to self pass in on att to $my_net no state flags any pass out on att from { $my _net self } no state flags any pass out on { $att_if0 $att_if1 } route-to { \ ($att_if0 $att_if0:peer) \ ($att_if1 $att_if1:peer) \ } round-robin from $my_net tag ROUTED ! tagged ROUTED \ no state flags any pass out on att to att:network this is on a multiple ATT T1 link, but it should work mostly the same. However, you probably won't have the :peer address and will have to specify the address. l8rZ, -- andrew - ICQ# 253198 - Jabber: [EMAIL PROTECTED] BOFH excuse of the day: Typo in the code
Re: Multipath to CISCO
On Thu, Nov 6, 2008 at 5:45 AM, andrew fresh [EMAIL PROTECTED] wrote: On Wed, Nov 05, 2008 at 09:40:02AM +, Stuart Henderson wrote: On 2008-11-05, Mikel Lindsaar [EMAIL PROTECTED] wrote: The other option I believe would be using PF to round robin the packets on both destinations using route-to rules. Would this work? it should, but you might need to make the rules stateless (no state). It works, and you do. SNIP EXAMPLE Thanks heaps Andrew. Mikel -- http://lindsaar.net/ Rails, RSpec and Life blog
Re: Multipath to CISCO
On Wed, Nov 05, 2008 at 01:02:33PM +1100, Mikel Lindsaar wrote: After reading http://www.openbsd.org/faq/faq6.html it looks that equal cost routing will not do what I want as it looks like each destination is mapped to one possible route out of a pool, which I believe means I'll only ever get 2mb/s per VOIP peer I connect to. AFAIK it's done by a hash from the source and destination pair. So, it depends on if you have multiple sources. But what do I know, I didn't read the source... The other option I believe would be using PF to round robin the packets on both destinations using route-to rules. Would this work? Why wouldn't it? At the ISP end we will be terminating into the back of a CISCO. The ISP is willing to work out what we need to make it work. I'm really trying to avoid having to buy an 1841. Don't have the budget or inclination to spend that much money for a little green box when I think my OpenBSD box can handle it. If you have two ethernets and you want to round-robin, trunk(4) might work too. -- Jussi Peltola