[j-nsp] Default OSPF Equal Cost Behaviour - EX Switches

[Paul Stewart]

Hi there....

 

I opened a JTAC ticket this evening and looking for some opinion - after I
closed the case I'm not sure of what I was told...

 

The ticket was opened because leaving a EX4200 VC there are two GigE
connections going to a Cisco 6500 - so equal cost OSPF.  According to the
routing table only one of those links was being selected (there was no
default per packet or per destination load balancing occurring).

 

The guy at JTAC suggested a policy as:

 

paul at dis1.lansdowne1> show configuration policy-options policy-statement
load-balancing

then {

    load-balance per-packet;

}

 

paul at dis1.lansdowne1> show configuration protocols ospf

export [ ospf_redistributes load-balancing ];

area 0.0.0.0 {

    interface vlan.130;

    interface vlan.904;

    interface vlan.971;

}

 

As there is quite a bit of voice traffic involved I asked if this was really
per packet or if it was "per flow" as he had mentioned.  I was told that
even though the configuration uses the wording "per-packet" that it's really
per flow occurring.  It does appear to actually be some kind of
source/destination hash similar or if not identical to default Cisco
behaviour ... can anyone confirm that for me?  I'd rather avoid per packet
load balancing if possible and stick to a round "round robin" per
destination format if possible...

 

It's running JTAC recommended JunOS 10.0R3.10

 

Thanks for your input,

 

Paul