n**********l
发帖数: 271 | 1 AS 65400 and AS65500 are peering with two links on the same router. How can
I use only one link and leave the other link as standby?
This is a customer & upstream scenario and there's nothing over the upstream
router. |
x*********n
发帖数: 28013 | 2 AS 65400 and AS65500 are customers and
peering with two links on the same router， this router is ISP？ |
n**********l
发帖数: 271 | 3 AS65400 customer
AS65500 upstream
BGP peering, one router at customer, one router at upstream
two circuits between routers
How to use only one link and leave the other as backup (use the backup link
only when the primary link goes down)
Why would anyone provision a circuit and only use it as backup? |
n**********l
发帖数: 271 | |
x*********n
发帖数: 28013 | 5 不太理解你的意思。
如果你是ISP，一个是customer，一个是你上流的ISP，那么你就是个transit，谈不上
什么load sharing。
比如
customer---circuit1--------YOU-------circuit2-------ISP
如果是
YOU---circuit1------ISP1
---circuit2-----ISP2，
那么就可以做load balance了。一般做50%，50%的比较容易，方案多。40%，60%，我不
太清楚怎么做。
link
【在 n**********l 的大作中提到】
: AS65400 customer
: AS65500 upstream
: BGP peering, one router at customer, one router at upstream
: two circuits between routers
: How to use only one link and leave the other as backup (use the backup link
: only when the primary link goes down)
: Why would anyone provision a circuit and only use it as backup?
|
x*********n
发帖数: 28013 | 6 这个例子，就是
CPE-------ISP1
-------ISP2
然后呢。用route-map 设metric
比如CPE有2个routes，然后呢，route1我设成next hop到ISP1，route2 设成ISP2，
这样，traffic就跟着什么routes走了。
这个route-map你还可以改成任何attributes，比如，local-preference，比如weight
，比如origin code。。功能是一样的。
一般来说用metric的很少，因为它的priority比较排后，而且实际情况metric不好算。
【在 n**********l 的大作中提到】
: http://www.onesc.net/communities/as701/bordergateproto.pdf
: Can I use MED(metric) in this scenario?
|
n**********l
发帖数: 271 | 7 CPE ---primary circuit------- ISP Router
---backup circuit -------
Only 1 ISP router, 1 CPE router involved. Backup circuit should be left
unused when the primary circuit is up. Graceful transition required.
Basically http://www.cisco.com/image/gif/paws/13762/40a.gif except one circuit should be utilized at any time. |
r********t
发帖数: 338 | 8 - eBGP peer with loopback (set ttl >1)
- create 2 static routes for peer loopback using each link.
- Use EEM to track link protocol of primary link.
Down event -> bring up secondary link
Up event -> shut secondary link
In reality, it is not a best practice for redundancy design.
【在 n**********l 的大作中提到】
: CPE ---primary circuit------- ISP Router
: ---backup circuit -------
: Only 1 ISP router, 1 CPE router involved. Backup circuit should be left
: unused when the primary circuit is up. Graceful transition required.
: Basically http://www.cisco.com/image/gif/paws/13762/40a.gif except one circuit should be utilized at any time.
|
n**********l
发帖数: 271 | 9 Thanks. I was thinking about some script to monitor link status and add/
remove the static route for the backup circuit interface (does not apply
because no control over upstream router).
Does it cause BGP restart? Won't the delay cause noticeable packet loss(
especially for 10G/40G links)?
【在 r********t 的大作中提到】
: - eBGP peer with loopback (set ttl >1)
: - create 2 static routes for peer loopback using each link.
: - Use EEM to track link protocol of primary link.
: Down event -> bring up secondary link
: Up event -> shut secondary link
: In reality, it is not a best practice for redundancy design.
|
r********t
发帖数: 338 | 10 Theoretically, BGP session should stay up if EEM flaps the interface quickly
enough.
However, it normally takes a couple seconds for EEM to react the event.
There will be a few seconds outage.
【在 n**********l 的大作中提到】
: Thanks. I was thinking about some script to monitor link status and add/
: remove the static route for the backup circuit interface (does not apply
: because no control over upstream router).
: Does it cause BGP restart? Won't the delay cause noticeable packet loss(
: especially for 10G/40G links)?
|
|
|
n**********l
发帖数: 271 | 11 What can I do on Juniper MX/Brocade XMR for the same purpose? Thanks. |
x*********n
发帖数: 28013 | 12 ip route xxxx xxxx xxxx
ip route xxxx xxxx xxxx 15
juniper和cisco都有graceful restart，hello packet里面含capability的，
一个router做另一个的restart trigger，一个要flap了，另一个存好备份随时准备。 |
n**********l
发帖数: 271 | 13 With static route both links will become active, the design requires one
link to serve as backup only. Is it?
【在 x*********n 的大作中提到】
: ip route xxxx xxxx xxxx
: ip route xxxx xxxx xxxx 15
: juniper和cisco都有graceful restart，hello packet里面含capability的，
: 一个router做另一个的restart trigger，一个要flap了，另一个存好备份随时准备。
|
x*********n
发帖数: 28013 | 14 这个叫floating routes，一个by default， static routes metric是0，一个是15.比
20小。记得我跟你说过的eBGP是20，OSPF是110，这些你面试很容易问的。
【在 n**********l 的大作中提到】
: With static route both links will become active, the design requires one
: link to serve as backup only. Is it?
|
x*********n
发帖数: 28013 | 15 给你分享几个题目，不是面试题。
1.router ospf 1，network xxxx command，和router bgp 100， network xx command
，这2个network command有什么区别？ |
n**********l
发帖数: 271 | 16
这个需要在upstream route上面也设置administrative distance的...这个CDN的
interview搞得我完全糊涂了....
【在 x*********n 的大作中提到】
: 这个叫floating routes，一个by default， static routes metric是0，一个是15.比
: 20小。记得我跟你说过的eBGP是20，OSPF是110，这些你面试很容易问的。
|
x*********n
发帖数: 28013 | 17 .......................
【在 n**********l 的大作中提到】
:
: 这个需要在upstream route上面也设置administrative distance的...这个CDN的
: interview搞得我完全糊涂了....
|
