OSPF stability

[Tom Harbert]

Hi,

I have two Ubuntu 14.04 physical machines running BIRD 1.4.5.

I am noticing the OSPF neighbor relationship is dieing.  it appears that
keepalives are not being received and the dead timer expires.

They are directly connected to each other using LACP.  There is no issue
with the physical connection and layer 3 connectivity remains up over the
interconnect when the problem occurs.  The problem is that OSPF is
advertising the loopback addresses used by iBGP which drop when the OSPF
fails.

2015-04-20 09:44:17 <TRACE> OSPF_1: Inactivity timer fired on interface
bond0 for neighbor x.x.x.x.
2015-04-20 09:44:17 <TRACE> OSPF_1: Neighbor x.x.x.x changes state from "
 full" to "    down".
2015-04-20 09:44:17 <TRACE> OSPF_1: Scheduling router-LSA origination for
area 0.0.0.0
2015-04-20 09:44:17 <TRACE> OSPF_1: Scheduling network-LSA origination for
iface bond0
2015-04-20 09:44:17 <TRACE> OSPF_1: Deleting neigbor.
2015-04-20 09:44:17 <TRACE> OSPF_1: New neighbor found: x.x.x.x on bond0
2015-04-20 09:44:17 <TRACE> OSPF_1: Neighbor x.x.x.x changes state from "
 down" to "    init".
2015-04-20 09:44:17 <TRACE> OSPF_1: Neighbor x.x.x.x changes state from "
 init" to "    2way".
2015-04-20 09:44:17 <TRACE> OSPF_1: Neighbor x.x.x.x changes state from "
 2way" to " exstart".
2015-04-20 09:44:19 <TRACE> OSPF_1: Originating router-LSA for area 0.0.0.0
2015-04-20 09:44:19 <TRACE> OSPF_1: Scheduling routing table calculation
2015-04-20 09:44:19 <TRACE> OSPF_1: Starting routing table calculation
2015-04-20 09:44:19 <TRACE> OSPF_1: Starting routing table calculation for
area 0.0.0.0
2015-04-20 09:44:19 <TRACE> OSPF_1: Starting routing table calculation for
inter-area (area 0.0.0.0)
2015-04-20 09:44:19 <TRACE> OSPF_1: Starting routing table calculation for
ext routes
2015-04-20 09:44:19 <TRACE> OSPF_1: Starting routing table synchronisation
2015-04-20 09:44:24 <TRACE> OSPF_1: Neighbor x.x.x.x changes state from "
exstart" to "exchange".
2015-04-20 09:44:26 <TRACE> OSPF_1: Neighbor x.x.x.x changes state from
"exchange" to " loading".
2015-04-20 09:44:27 <TRACE> OSPF_1: Neighbor x.x.x.x changes state from "
loading" to "    full".
2015-04-20 09:44:27 <TRACE> OSPF_1: Scheduling router-LSA origination for
area 0.0.0.0
2015-04-20 09:44:27 <TRACE> OSPF_1: Scheduling network-LSA origination for
iface bond0
2015-04-20 09:44:27 <TRACE> OSPF_1: Scheduling routing table calculation

Configuration:

# OSPF process 1
protocol ospf OSPF_1 {
description "OSPF area 0";
area 0 {
interface "bond0" { # To x-x-x-01
                        hello 1;
                        dead count 4;
                        type ptp;
cost 1;
                };
stubnet x.x.x.x/29 { # bond2.51
cost 0;
};
stubnet x.x.x.x/32 { # lo:1
cost 0;
};
                stubnet x.x.x.x/31 {     # bond1.31
                        cost 0;
                };
                stubnet x.x.x.x/31 {     # bond1.33
                        cost 0;
                };
        };
      #import filter import_OSPF;
        #export filter export_OSPF;
}


Am I doing anything out of the ordinary here?  Has anyone else had problems
with OSPF dropping, this happens 5-10 times a day.  I see no issues with
the CPU utilisation on either of these hosts.

I have used workarounds such as temporary static routes and also found that
adjusting the dead count from 4 to 8 has kept it stable for ~2 days.

2015-04-20 10:26:53 <TRACE> OSPF_1: Changing dead interval on interface
bond0 from 4 to 8

Thanks,
Tom.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://trubka.network.cz/pipermail/bird-users/attachments/20150421/c8977572/attachment.html>