Support Center > Search Results > SecureKnowledge Details
When the same OSPF interface is disconnected on both cluster members, OSPF Hello messages might not be sent on the other OSPF interfaces Technical Level
Symptoms
  • In ClusterXL, when multiple interfaces are used for OSPF, and the same interface is disconnected on both cluster members, OSPF Hello messages might not be sent on the other OSPF interfaces from the member that is in 'Active Attention' state.

  • Example from OSPF trace:

    ........................
    Aug 26 15:42:25 task_send_packet_iovec: task OSPF2_IO socket 18 length 72 if_index 16 to 224.0.0.5
    Aug 26 15:42:25 OSPFv2: SENT 192.168.135.1 -> 224.0.0.5 Hello length 52 (52)
    rtrid 192.168.55.254 backbone csum 0x8101 authtype Simple
    ASE mask 255.255.255 hello 1 pri 0 dead 4
    dr 10.203.135.4 bdr 10.203.135.5 nbrs 10.205.0.1 10.205.0.2
    OSPF SENT end of packet
    Aug 26 15:42:25
    Aug 26 15:42:25 OSPF CLUSTER: iface 192.168.135.1(bond0.3335) send crypto sequence <0>
    Aug 26 15:42:25 task_timer_dispatch: returned from OSPF2_default_Hello, rescheduled in 0.999

    Aug 26 15:42:25 CLUSTER: Proto 7 blocks sending in cluster
    Aug 26 15:42:25 OSPF: interface 192.168.27.254(bond1.403) DOWN

    Aug 26 15:42:26 main: running normal priority timer queue
    Aug 26 15:42:26 task_timer_dispatch: calling OSPF2_default_Hello, late by 0.000
    Aug 26 15:42:26 task_timer_dispatch: returned from OSPF2_default_Hello, rescheduled in 0.999
    Aug 26 15:42:26 main: ran 1 timer
    ........................
Cause

When an interface associated with OSPF goes down, RouteD disables sending OSPF cluster messages in anticipation of a failover.

As a result, if using multiple OSPF interfaces, and one OSPF interface is disconnected on both members, both members will be blocked for sending OSPF messages, which will disrupt all the OSPF sessions, not just the one on the disconnected interface.

When using short Hello/Dead intervals (such as 1 sec for Hello / 4 sec for Dead), the blocking may prevent sending an Hello for 4 sec until the cluster failover occurs.


Solution
Note: To view this solution you need to Sign In .