Re: [c-nsp] Change BGP default-originate to IGP?
Thanks for the tips, we'll have a play with some of the options suggested around originating the default. On 05/10/2012, at 11:52 AM, Anton Kapela wrote: also +1 to inter-border router ibgp sessions over some other layer2 path/port pair/etc -- one should always have that, unless you can't for some strange reason. On 27/09/2012, at 11:24 PM, David Prall wrote: As well could put a GRE Tunnel or VLAN between the two ASR's and run iBGP between the two. You control the path between the two routers, so the tunnel can be over a jumbo frame capable path. I'm glad a iBGP session between the ASRs over a GRE tunnel was mentioned, as that's exactly what we have running and I was questioning whether this was a bad practice or not... Thanks, Tom ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] Change BGP default-originate to IGP?
-Original Message- From: Tom Lanyon [mailto:tom+c-...@oneshoeco.com] I'm glad a iBGP session between the ASRs over a GRE tunnel was mentioned, as that's exactly what we have running and I was questioning whether this was a bad practice or not... Thanks, Tom [dprall] It's the Duct Tape of Networking David -- http://dcp.dcptech.com ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] Change BGP default-originate to IGP?
On Thu, Sep 27, 2012 at 8:56 PM, Tom Lanyon tom+c-...@oneshoeco.com wrote: On 28/09/2012, at 4:03 AM, Christian Meutes wrote: The best way is here really to inject the defaults via ISIS. Great - that's my current thinking too. I just wanted to double-check that this was reasonable and I wasn't missing something! Another option -- manually delay originating any bgp routes from the ASR's to the 4900's with adjusted MRAI, to something, say 1.5x longer than 'usual case' convergence: http://www.cisco.com/en/US/docs/ios/iproute_bgp/command/reference/irg_bgp3.html#wp1105562 also +1 to inter-border router ibgp sessions over some other layer2 path/port pair/etc -- one should always have that, unless you can't for some strange reason. -Tk ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] Change BGP default-originate to IGP?
So if I understood it correctly you are concerned that the router will start to originate the default prior to receiving full BGP table from its upstream right? The simplest solution would be to place a static default route pointing to the upstream -so in case of the above happens the router wouldn't blackhole till it gets the full feed. Or you can base the condition of default advertisement on several prefixes from all around the place instead of just one -but trying to match e.g. 10 prefixes from 400k well... The problem with this though is that on IOS you can't AND the match ip address statements under the route-map only OR -applies to all the match statements of the same kind. So you could use static routes for the list of desired prefixes each with a high AD -redistribute them into bgp with specific community and match for the prefixes+community in non-exist map. So once you'll get all the prefixes from BGP -only after all cease to exist in the bgp table with the specific community of yours -the router will start to advertise the default route. adam -Original Message- From: cisco-nsp-boun...@puck.nether.net [mailto:cisco-nsp-boun...@puck.nether.net] On Behalf Of Tom Lanyon Sent: Thursday, September 27, 2012 4:58 AM To: David Prall; cisco-nsp Subject: Re: [c-nsp] Change BGP default-originate to IGP? Is there a specific order in which BGP updates are sent/exchanged/processed? The concern I have with tracking upstream routes is that the route tracked would need to be one of the last routes received (if not the last) to ensure that the router has full visibility. This seems quite non-deterministic and so potentially fraught with 'weirdness'. Thanks, Tom On 27/09/2012, at 12:19 PM, David Prall wrote: Why not use selective advertisement of the default based on receiving a specific route from your carrier or an upstream you know to be stable. http://www.cisco.com/en/US/docs/ios/12_3/iproute/command/reference/ip2 _n1g.h tml#wp1037042 David ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/ ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] Change BGP default-originate to IGP?
Another option is to lock the static default route to a tracked object, and based that object on eem scripting or something similar which has a built in delay after the bgp peer reaches X prefix count. It's definitely more creaky though. Something you should keep in mind is you're going to have some blackholing regardless of what you do, due to the return path experiencing the same issue as it propagates out. It's just a question of what levels you are comfortable with. -Blake On Thu, Sep 27, 2012 at 3:40 AM, Adam Vitkovsky adam.vitkov...@swan.skwrote: So if I understood it correctly you are concerned that the router will start to originate the default prior to receiving full BGP table from its upstream right? The simplest solution would be to place a static default route pointing to the upstream -so in case of the above happens the router wouldn't blackhole till it gets the full feed. Or you can base the condition of default advertisement on several prefixes from all around the place instead of just one -but trying to match e.g. 10 prefixes from 400k well... The problem with this though is that on IOS you can't AND the match ip address statements under the route-map only OR -applies to all the match statements of the same kind. So you could use static routes for the list of desired prefixes each with a high AD -redistribute them into bgp with specific community and match for the prefixes+community in non-exist map. So once you'll get all the prefixes from BGP -only after all cease to exist in the bgp table with the specific community of yours -the router will start to advertise the default route. adam -Original Message- From: cisco-nsp-boun...@puck.nether.net [mailto:cisco-nsp-boun...@puck.nether.net] On Behalf Of Tom Lanyon Sent: Thursday, September 27, 2012 4:58 AM To: David Prall; cisco-nsp Subject: Re: [c-nsp] Change BGP default-originate to IGP? Is there a specific order in which BGP updates are sent/exchanged/processed? The concern I have with tracking upstream routes is that the route tracked would need to be one of the last routes received (if not the last) to ensure that the router has full visibility. This seems quite non-deterministic and so potentially fraught with 'weirdness'. Thanks, Tom On 27/09/2012, at 12:19 PM, David Prall wrote: Why not use selective advertisement of the default based on receiving a specific route from your carrier or an upstream you know to be stable. http://www.cisco.com/en/US/docs/ios/12_3/iproute/command/reference/ip2 _n1g.h tml#wp1037042 David ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/ ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/ ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] Change BGP default-originate to IGP?
As well could put a GRE Tunnel or VLAN between the two ASR's and run iBGP between the two. You control the path between the two routers, so the tunnel can be over a jumbo frame capable path. I would still do a selective advertisement for the default, so that traffic doesn't need to traverse the both routers, but this would stop the blackholes for routes unlearned. Especially if it is the upstream that is having issues, instead of the direct connection. David -- http://dcp.dcptech.com -Original Message- From: Adam Vitkovsky [mailto:adam.vitkov...@swan.sk] Sent: Thursday, September 27, 2012 4:41 AM To: 'Tom Lanyon'; 'David Prall'; 'cisco-nsp' Subject: RE: [c-nsp] Change BGP default-originate to IGP? So if I understood it correctly you are concerned that the router will start to originate the default prior to receiving full BGP table from its upstream right? The simplest solution would be to place a static default route pointing to the upstream -so in case of the above happens the router wouldn't blackhole till it gets the full feed. Or you can base the condition of default advertisement on several prefixes from all around the place instead of just one -but trying to match e.g. 10 prefixes from 400k well... The problem with this though is that on IOS you can't AND the match ip address statements under the route-map only OR -applies to all the match statements of the same kind. So you could use static routes for the list of desired prefixes each with a high AD -redistribute them into bgp with specific community and match for the prefixes+community in non-exist map. So once you'll get all the prefixes from BGP -only after all cease to exist in the bgp table with the specific community of yours -the router will start to advertise the default route. adam -Original Message- From: cisco-nsp-boun...@puck.nether.net [mailto:cisco-nsp-boun...@puck.nether.net] On Behalf Of Tom Lanyon Sent: Thursday, September 27, 2012 4:58 AM To: David Prall; cisco-nsp Subject: Re: [c-nsp] Change BGP default-originate to IGP? Is there a specific order in which BGP updates are sent/exchanged/processed? The concern I have with tracking upstream routes is that the route tracked would need to be one of the last routes received (if not the last) to ensure that the router has full visibility. This seems quite non-deterministic and so potentially fraught with 'weirdness'. Thanks, Tom On 27/09/2012, at 12:19 PM, David Prall wrote: Why not use selective advertisement of the default based on receiving a specific route from your carrier or an upstream you know to be stable. http://www.cisco.com/en/US/docs/ios/12_3/iproute/command/reference/ip2 _n1g.h tml#wp1037042 David ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/ ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] Change BGP default-originate to IGP?
The best way is here really to inject the defaults via ISIS. On 27.09.2012, at 04:04, Tom Lanyon tom+c-...@oneshoeco.com wrote: Hi list, In an enterprise network I have a core of 4900Ms with a few ASR1ks hanging off to handle upstream connectivity. As an example: Upstream1 - [ASR1k]--[4900M]--[4900M]--[ASR1k] - Upstream2 || || ServersWorkstations, etc The ASRs and 4900Ms are running BGP and ISIS with full tables on the ASRs and mostly just defaults on the 4900Ms. The ASRs are originating defaults via BGP but on reload they are blackholing traffic whilst BGP converges. I've seen that OSPF and ISIS have 'wait-for-bgp' overload bits available and have been questioning whether switching to an IGP-generated default with wait-for-bpg is the correct solution here. Any thoughts? Thanks, Tom ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/ ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] Change BGP default-originate to IGP?
On 28/09/2012, at 4:03 AM, Christian Meutes wrote: The best way is here really to inject the defaults via ISIS. Great - that's my current thinking too. I just wanted to double-check that this was reasonable and I wasn't missing something! Tom ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
[c-nsp] Change BGP default-originate to IGP?
Hi list, In an enterprise network I have a core of 4900Ms with a few ASR1ks hanging off to handle upstream connectivity. As an example: Upstream1 - [ASR1k]--[4900M]--[4900M]--[ASR1k] - Upstream2 || || ServersWorkstations, etc The ASRs and 4900Ms are running BGP and ISIS with full tables on the ASRs and mostly just defaults on the 4900Ms. The ASRs are originating defaults via BGP but on reload they are blackholing traffic whilst BGP converges. I've seen that OSPF and ISIS have 'wait-for-bgp' overload bits available and have been questioning whether switching to an IGP-generated default with wait-for-bpg is the correct solution here. Any thoughts? Thanks, Tom ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] Change BGP default-originate to IGP?
Why not use selective advertisement of the default based on receiving a specific route from your carrier or an upstream you know to be stable. http://www.cisco.com/en/US/docs/ios/12_3/iproute/command/reference/ip2_n1g.h tml#wp1037042 David -- http://dcp.dcptech.com -Original Message- From: cisco-nsp-boun...@puck.nether.net [mailto:cisco-nsp-boun...@puck.nether.net] On Behalf Of Tom Lanyon Sent: Wednesday, September 26, 2012 10:04 PM To: cisco-nsp Subject: [c-nsp] Change BGP default-originate to IGP? Hi list, In an enterprise network I have a core of 4900Ms with a few ASR1ks hanging off to handle upstream connectivity. As an example: Upstream1 - [ASR1k]--[4900M]--[4900M]--[ASR1k] - Upstream2 || || ServersWorkstations, etc The ASRs and 4900Ms are running BGP and ISIS with full tables on the ASRs and mostly just defaults on the 4900Ms. The ASRs are originating defaults via BGP but on reload they are blackholing traffic whilst BGP converges. I've seen that OSPF and ISIS have 'wait-for-bgp' overload bits available and have been questioning whether switching to an IGP-generated default with wait-for-bpg is the correct solution here. Any thoughts? Thanks, Tom ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/ ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
Re: [c-nsp] Change BGP default-originate to IGP?
Is there a specific order in which BGP updates are sent/exchanged/processed? The concern I have with tracking upstream routes is that the route tracked would need to be one of the last routes received (if not the last) to ensure that the router has full visibility. This seems quite non-deterministic and so potentially fraught with 'weirdness'. Thanks, Tom On 27/09/2012, at 12:19 PM, David Prall wrote: Why not use selective advertisement of the default based on receiving a specific route from your carrier or an upstream you know to be stable. http://www.cisco.com/en/US/docs/ios/12_3/iproute/command/reference/ip2_n1g.h tml#wp1037042 David ___ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/