[ixpmanager] LAG and SFLOW workaround

Nick Hilliard nick at inex.ie
Thu Feb 6 00:31:26 GMT 2014


On 05/02/2014 23:21, Brian Thompson wrote:
> Could you describe "foundation work" and how I might leverage that, to make
> sure I am moving the ball in the right direction?

Years ago we used cisco equipment, so we created the notion of a
virtualinterface table which would be an abstraction for handling LAGs of
all forms.  When you edit a virtual interface, you can see that it has one
or more physical interfaces associated with it, and it also will support
one or more vlan interfaces.  There's also a hidden option to configure a
Virtual Interface Name; the intention was to use this for "ae0" or
"PortChannel1" or whatever.

In order to support named virtual lag interfaces, we would need to:

- require the "Virtual Interface Name" (click on "Advanced Options" in the
virtual interface edit page to see this) to be filled in for certain types
of switches.

- not use this virtual interface name when polling for bitcount graphs
(mrtg, etc)

- use this virtual interface name when polling switches for layer 2 mac
address -> port mappings (update-l2database)

- use this virtual interface name as an endpoint in sflow-to-rrd-handler.

It's not weeks of work or anything, but it's something that is troublesome
to do without having access to juniper boxes to test it on because you end
up coding in the dark.

It would be useful if you could post some sample sflowtool output on gist?
 Something like:

# sflowtool -4 -l | grep <mac address of router on LAG port> | head -50

This would give us some data to work on.

> I also am
> not familiar with 'kit' could you send me a reference so I can put more
> context around your comments.

oops, local idiom.  Kit == equipment.

Nick




More information about the ixpmanager mailing list