[ixpmanager] Disparity between MRTG and SFLOW graphs.

[Nick Hilliard]

Lolade Olori wrote on 22/07/2021 13:51:
> For Eg, a sensor which has 300Mbps and 4Gbps of input and output traffic 
> respectively on MRTG will be around 0Mbps and 100Mbps on Sflow.

There are a several things that could cause this:

- ports not configured as type: Peering
- switch ports not configured to have ingress sflow enabled on each port
- switches which are configured to rate-limit the sampling rate
- mac address database not configured correctly or not properly matching 
reality

Generally, sflow is one of the more difficult aspects of IXP Manager to 
get (and stay) working correctly because there are a lot of moving parts 
and interactions with non-ixp-manager components.

As a check-list, you should ensure:

- all the ports used by IXP participants are configured as type: Peering

- manually check that the port configuration on your switches includes 
sflow.

- make sure that the overall sample rate doesn't exceed the switch's 
default sflow sample export rate

- make sure that all MAC addresses are correctly configured in the "MAC 
Addresses | Configured Addresses" list.

- Check that each ixp participant port has a mac address associated with 
it in "update-l2database.pl --debug"

Nick