Junos OS

Hello,
I have a Juniper M120 router running JUNOS 9.6R1.13. The SNMP trapping behavior on this device seems to differ from the Cisco routers I've worked on when it comes to the bgpBackwardTransition trap.

On a Cisco, when a BGP session is dropped, it will send one bgpBackwardTransition trap. Upon it's reconnection, it will then transmit a bgpEstablished trap.

On the Juniper, it will transmit a bgpBackwardTransition trap when the session is dropped. However, where it differs is that it will re-transmit the trap each time it cycles the BGP states in an attempt to re-connect the downed session.

Is this intended functionality? Has anyone else experienced this?

Cheers,
Christoph

---

On the Juniper, it will transmit a bgpBackwardTransition trap when the session is dropped. However, where it differs is that it will re-transmit the trap each time it cycles the BGP states in an attempt to re-connect the downed session.

Is this intended functionality? Has anyone else experienced this?

---

Yes. This is intended.  This behavior matches the description in the BGP4-MIB for this oid.  There are intermediate states in the FSM (Finite State Machine) between Idle and Established. The peer FSM is oscillating, due to timer expiry, receiving an error or notification, etc.  This causes the trap to be sent each time the FSM goes from a higher, to a lower state.

 

/* from RFC 4273 */

 

"bgpBackwardTransNotification NOTIFICATION-TYPE
    OBJECTS { bgpPeerRemoteAddr,
              bgpPeerLastError,
              bgpPeerState      }
    STATUS  current
    DESCRIPTION
            "The bgpBackwardTransNotification event is
             generated when the BGP FSM moves from a higher
             numbered state to a lower numbered state.
              This Notification replaces the
             bgpBackwardsTransition Notification."
    ::= { bgpNotification 2 }"

 

Regards,

Ben

 

Is there any way with JUNOScript or otherwise to have the functionality line up with how Cisco does it? While I'm usually a fan of conforming to established standards, we have a specificly designed application that doesn't work well with the repeating bgpBackwardTransition traps.

You could use event-options to monitor RPD messages regarding BGP state changes and create a trap using these messages.  The RPD message only shows up when a BGP neighbor goes to or from the Established state.  You would need to use the JUNIPER-SYSLOG-MIB for the trap.   This message will show up the messages log file if your syslog level is set to warning.  

 

http://www.juniper.net/techpubs/software/junos/junos92/syslog-messages/html/rpd8.html

 

> show configuration event-options

policy rpd_bgp {
    events rpd_bgp_neighbor_state_changed;
    then {
        raise-trap;
    }
}

Dv8 provides a good suggestion.  Make sure to configure the bgp sessions to log the state transitions with the following.

 

log-updown

 

Regards,

Ben

 

dv8:

This worked like a charm! I'm getting traps only when the session establishes and gets torn down.

 

Thank you for your assistance!

 

I actually created the event-options policy as stated in the earlier post. I also have BGP groups with each group having the "log-updown" command. But does not work for me.

 

I am applying this on a J2320. Do I have to apply the event-option policy anywhere in the config?