Hi everyone! I'm the sole network engineer recently hired to work on a fairly new MPLS network. It was designed and handed over to my employer before I came on board. I've begun testing the network and I've noticed a potentially huge flaw in the design.
While this is a fairly straight forward MPLS network, it does have a slight hitch. All L1 connectivity is done via microwave which connects to around 100 sites. Every site has a 3560 switch (L2 ONLY) which was meant to serve as the PE and all hub sites have Cisco ASR1002s. Due to being microwave, a lot of the sites are in remote locations and the microwave paths are crossed over with the 3560. This network is doing iBGP in the core with OSPF as the IGP. So if I were to bring on a customer at SW1 in the middle of a segment, I'd peer them via eBGP with R2 and R3 with the intent that if the link failed, it would go in the other direction to reach CE2.
Here is a sample segment from my lab:
Although, there are only 2 switches in the drawing, sometimes there are as many as 6 3560 switches between ASRs.
The issue I'm having occurs when the microwave path gets interrupted between switch sites. As depicted in the drawing, when I shut down SW1 port F0/7, CE1 can no longer reach CE2.
The issue is that R2, is still preferring its 0.0.0.0 path to CE1 because it acts like it isn't aware the path between SW1 and SW2 has failed.
Is there something I could do with BGP to make it not prefer its local originated route?
I realize since it's Cisco I could just change the weight something higher than 32768 on the R3 neighbor, but when I did this in my lab, it only went to R3 and still wouldn't failover.
I hope bringing 3560s into the iBGP mesh isn't the only way to fix this.
I'm sorry about the lengthy post and I'm truly appreciative any input I get. Thanks!
No comments:
Post a Comment