Hey guys,
What are your feelings about this? I've worked with people that like to redistribute connected with a route map, and seen docs that say to use network statements instead - though now I can't remember why... think it had something to do with reducing workload on the router? Anyway - what are your thoughts? Pros? Cons?
EDIT - One thing I remember mentioned as a pro to redistributing w/route map was that it was less typing. In the thread I saw, that was debunked saying that each route would have to be added to the route map anyway so it doesn't save you that much. I've experienced this myself, and especially in a dual stack environment I'm setting up - getting a little complex with the multiple matches for v4 and v6. Though I will say it looks pretty clean in the router bgp config section.
I'd use a prefix list rather than a route map or network statement. due to the more granular control
route map if I wanted to tag a route.
Quote from: ristau5741 on March 26, 2015, 12:34:49 PM
I'd use a prefix list rather than a route map or network statement. due to the more granular control
route map if I wanted to tag a route.
How do you use prefix lists by themselves for that? Wouldn't you need to configure those prefix lists as match statements under a route map?
you can use prefix lists directly in a BGP neighbor statement.
here is a good reference. (compares ACL's and prefix lists), but shows how it's done. just skip down to the prefix list section after looking a the initial configs. (or not, the ACL section is not very long)
https://mellowd.co.uk/ccie/?p=447
BTW: am I the first to reference Darren's Blog on the new site? :)
Quote from: ristau5741 on March 26, 2015, 02:42:44 PM
you can use prefix lists directly in a BGP neighbor statement.
here is a good reference. (compares ACL's and prefix lists), but shows how it's done. just skip down to the prefix list section after looking a the initial configs. (or not, the ACL section is not very long)
https://mellowd.co.uk/ccie/?p=447
BTW: am I the first to reference Darren's Blog on the new site? :)
Ahhhhhh "distribute-list" - forgot about that one. Been a while since I've used IOS in that capacity. Thanks. Wouldn't you still need to use network statements with that though to actually advertise the subnets, as I believe Darren does in his blog?
In general, I hate redistribution. It might seem innocent when first implemented but it can bite you in the ass later.
My stance is to never use it unless you have no other choice.Even then tie a route-map to it ALWAYS!
I've always done it with prefix lists and route maps. I worked at an ISP for awhile and this became second nature. My new location redistributes almost everywhere with no filtering. I've seen first hand what that can do and have been working to fix it. Filter, Filter, Filter!
In BGP redistributing means a different origin code. This has implications for the best path algorithm.
I agree with everyone saying there is no good reason aside from laziness not to guard it with a prefix list or route-map.
And no ristau I referenced it when I asked a juniper MPLS PE-CE OSPF routing question :)
Depends on what you're using BGP for... On an internet routes with public AS and IP range, network statement and prefix-list.
On a PE or CE router in MPLS: redistribute and route-maps.
any particular reason to use redistribute over network to advertise a directly connected range on a MPLS PE?
Well, for a private MPLS: upstream CE uses a dynamic routing protocol like OSPF. So adding a new subnet in the OSPF does not require config changes on the PE.
oh you mean redistributing from the PE-CE protocol. Yeah.
I thought you were saying you'd redistribute the connected subnet (the /30 or /29 between PE and CE) into the VRF BGP instead of using network, for a specific reason.
Eh, that too.
I worked with a private MPLS cloud, managing CE, PE and P, and on-site routing on several sites. It was all a large intranet zone without firewalling in between because it was a homogenous network. Redistribution everywhere. It was easy. However, for a 'real' service provider that may be different.
Quote from: that1guy15 on March 26, 2015, 11:14:08 PM
In general, I hate redistribution. It might seem innocent when first implemented but it can bite you in the ass later.
My stance is to never use it unless you have no other choice.Even then tie a route-map to it ALWAYS!
So in large environments you'd have a really long BGP config utilizing network statements?
Quote from: AspiringNetworker on March 31, 2015, 11:31:03 AM
Quote from: that1guy15 on March 26, 2015, 11:14:08 PM
In general, I hate redistribution. It might seem innocent when first implemented but it can bite you in the ass later.
My stance is to never use it unless you have no other choice.Even then tie a route-map to it ALWAYS!
So in large environments you'd have a really long BGP config utilizing network statements?
I guess "never" was the wrong way to phrase this. When I wrote this Im pretty sure I was talking about all redistribution in general not just IGP->BGP. Also think I just came off a mean INE redistribution lab...
In "most" standard BGP situations I would lean towards network statement or aggregate addresses to keep the BGP config simple. But in a situation with a large number of prefixes that cant be aggregated then sure it will be cleaner to place them all under a route-map/prefix-list and redistribute. MPLS VPN sure redistribute all the time.
But just like always I think it depends on your situation and what functionality you want. Redistribution allows you to dynamically add/remove prefixes via your IGP where-as with network statements its back to a manual process unless your subnets fit within an aggregate.
Quote from: that1guy15 on March 31, 2015, 01:12:13 PM
Quote from: AspiringNetworker on March 31, 2015, 11:31:03 AM
Quote from: that1guy15 on March 26, 2015, 11:14:08 PM
In general, I hate redistribution. It might seem innocent when first implemented but it can bite you in the ass later.
My stance is to never use it unless you have no other choice.Even then tie a route-map to it ALWAYS!
So in large environments you'd have a really long BGP config utilizing network statements?
I guess "never" was the wrong way to phrase this. When I wrote this Im pretty sure I was talking about all redistribution in general not just IGP->BGP. Also think I just came off a mean INE redistribution lab...
In "most" standard BGP situations I would lean towards network statement or aggregate addresses to keep the BGP config simple. But in a situation with a large number of prefixes that cant be aggregated then sure it will be cleaner to place them all under a route-map/prefix-list and redistribute. MPLS VPN sure redistribute all the time.
But just like always I think it depends on your situation and what functionality you want. Redistribution allows you to dynamically add/remove prefixes via your IGP where-as with network statements its back to a manual process unless your subnets fit within an aggregate.
Gotcha - thanks for the tips.