MTU is standart 1500. I'm just realized that i forgot to explain network topology, so, if it helps: I have two routers R1C1 and R2C1 with interfaces bond0 (eth0 + eth1). On both routers i setting up many VLAN (about 250) - for example bond0.001-bond0.250 with addresses 10.16.a.1/24 and 10.16.a.2/24. Over this VLAN works VRRP (keepalived) with VIP 10.16.a.254. For one half of interfaces default router (VRRP address owner) is R1C1 and for another half R2C1 This interfaces I add to Area 1 and make them stub.
Between this routers I have two VLAN: 4012 and 4021 also in Area 1 for interconnection of this routers. On every router i have VLAN for connecting them to Core router in Area0 (Vlan 4001 for R1C1 and 4002 for R2C1). Core announce to R1C1 and R2C1 default route and 10.0.0.0/8. Also i have Router ID addresses on lo - 10.0.1.1/32 and 10.0.1.2/32 in Area 0. If I setup clients vlan (bond0.001-bond0.250) only on one of the routers (R1C1 or R2C1, no matter on what, for example R2C1, like in pastebin), and enable second, segfault occurs right after state FULL on first router who try to announce prefixes to his neighbour R1C1, who have no clients interfaces up. Core router in Area0 always receive routes without any trouble. I discovered three ways to avoid segfault, and make things works: 1) disable interfaces 4012 and 4021 which interconnects routers R1C1 and R2C1. In that case another router receive routes from core, which receive it from first, so traffic betwen routers will go through Core router, not wery good. 2) set interfaces 4012 and 4021 in Area 0, I think its wrong topology. 3) set MTU on interfaces 4012 and 4021 to 9000. From debug i discovered that LS Update massage have size about 6500. I hope that helps. Another question a wanted to ask - I found that in routing table appears routes to VRRP IP (10.16.a.254), so, in result, i have in routing table route to network 10.16.a.0/24 and route to host in this network 10.16.a.254. Is it possible to avoid OSPF to learn secondary addresses /32 from interface?
