0

I want to use BGP failover on three VMs. I installed a BGP daemon (BIRD) on the local VMs to achieve this and created a NIC with the floating IP: eth0:0.

However, I cannot 'up' the network interface on all VMs at the same time, but that is the behavior I need for BGP failover.

I get the following error:

[root@proxy2 network-scripts]# ifup eth0:0
ERROR     : [/etc/sysconfig/network-scripts/ifup-eth] Error, some other host (xxx) already uses address xxx.

How can I disable this check?

William
  • 584
  • 2
  • 9
  • 27

2 Answers2

0

Add the line

ARPCHECK="no"

to the /etc/sysconfig/network-scripts/ifcfg-* files of each interface that needs to have an overlapping address.

telcoM
  • 87,318
  • 3
  • 112
  • 232
0

Do not use the same address in several actual NIC (sub)interfaces.

Define different IP addresses, and use the dummy/localhost interfaces for VIP addresses, with preference for a different RFC 1918 netblock. Announce the routes in BGP. That how is usually done, even in Cisco. Reading some material in the CCNA+CCNP courses helps too.

See OSPF route costs in BIRD

and OSPF: Migrating Quagga to BIRD

and

OSPF import route filter for BIRD

While they are dealing with OSPF, they should give you an idea how to deal with the floating IP addresses.

See also How can we create multiple dummy interfaces on Linux?

See also Anycast DNS - Part 5, Using BGP

Rui F Ribeiro
  • 55,929
  • 26
  • 146
  • 227
  • Failover with BGP is fine. – William Dec 16 '17 at 10:26
  • @WilliamD.Edwards The way to do it cleanly is with loopback/dummy interfaces and providing routing for them over the official IP addresses. Plus, often they are put on separate networks for additional redundancy. And this way you do need the ARPCHECK "hack". I have several anycast BGP DNS servers working for years with this method. – Rui F Ribeiro Dec 16 '17 at 10:37
  • Yes, it is the cleanest method. – William Dec 16 '17 at 10:46
  • (correction, this way you do NOT need the ARPCHECK hack). – Rui F Ribeiro Dec 16 '17 at 10:48