We faced very strange issue with our linux router. We have CentOS 6.3 server with several VLANs on it. Each VLAN interface has two IP, one management IP for ucarp failover checks and one virtual IP to share with second server. You can see schema below:
[ ] 172.16.0.100 --> 172.16.0.1 [Virtaul IP VLAN 10 ] [ Router ] [ Virtual IP VLAN 20 ] 172.16.1.1 --> 172.16.1.[2-4] 172.16.0.7 [Management IP VLAN 10 ] [ ] [ Management IP VLAN 20 ] 172.16.1.7
So the problem is that sometimes traffic from one network can't reach ip address from other net from specific diapason 172.16.1.[2-4] (/30 network). All other addresses from vlan20 are ok. It can be fixed with ifdown/ifup vlan20, but it is very very strange and rediculous bihaviour for router. Also from router ping -I 172.16.0.7 172.16.1.2 is ok, but ping -I 172.16.0.1 172.16.1.2 is dead. I can see icmp requests and responses on 172.16.1.2, and I can see them on router, but ping can't get it. Even more ping -I 172.16.0.1 172.16.1.20 is ok.