We want to connect one of the CE routers to two our PE's (active/standby, no loadbalancing) and use dynamic protocol for route exchange.
Both PE's and CE share the same subnet.
What are our options? Can we use HSRP, make one PE active and align its BGP configuration accordingly?
Or is some other method more preferred? Like advertising routes to CE with higher metric from the less preferred PE, or something..
Couple of questions -
1) are you using BGP between CE and PE's.
2) you have control of your PEs ie.. do you configure them
3) is IBGP running between peers.
if you use HSRP that will only effect outbound traffic so return could come from either PE unless you modify the BGP advertisements from the PEs to the rest of the MPLS WAN.
In addition if you use a dymanic routing protocol betweem CE and PEs then this will override HSRP ie. HSRP would only really work if you manually pointed the CE to the HSRP active device.
1) BGP is used for PE-CE route exchange
2) Yes we have full control of PE
3) PE run iBGP between them via MPLS core
Any suggestion of what is the best way to achieve desired active/standby PE is welcome. Whether it involves HSRP or not is not a big deal (HSRP just came to my mind first).
If you only have one CE then i would use the BGP weight attribute which is Cisco specific. So you receive routes from both PEs and you simply assign a higher weight to the routes received from your primary PE.
The weight attribute is the first thing considered in the BGP path selection on Cisco routers and it is only significant on the router itself ie. it is not passed to any other BGP routers.
Routes learnt from EBGP peers are 0 by default. The higher the weight the more preferred the route is.
That will take care of outgoing traffic. Are you concerned about incoming as well ?
I have one seemingly interesting point here, when the active's PE interface to core goes down then the CE's weight will be still active pointing to this now faulty PE, until tables are flushed on PE. Is this correct?
And yes returning traffic should be symmetrical, via the same active PE. I guess MED is the way here.
If the primary PE link to the MPLS core goes down then it won't be advertising remote site routes which leads the CE to go through the other eBGP leg.
for inbound traffic, yes MED is the way to go.
One last point be aware when you advertise your LAN routes via the CE( static routes or IGP) not to cause any recursive lookup issues.
You can play around with the BGP timers to speed up the detection of the loss of an EBGP neighbor. As you control both the CE and the PEs you can look to set quite aggressive timers.
Agree that MED is the way to go for inbound traffic.