Monitor packet loss on an interface

Question

I am trying to create a monitor/alert for our SD-WAN devices (CloudGenix 2k and 3k) on the interface where the carrier circuit terminates. I have been advised by CGX support to use the ifInDiscards/ifOutDiscards OIDs, but they haven't been very useful.

From my understanding, the high packet loss only monitors at the node level, not on individual interfaces.

Really, all i'm looking for here is some way to monitor and alert on the carrier circuit when there is an issue on these devices. Has anyone done this before? Any help would be much appreciated.

rmullal · Answer

we're specifically looking for packet loss on the port where the carrier circuit terminates. basically we want to be able to monitor the carrier circuit, and this is the only thing i could come up with. i'm working on creating the UnDP now for errors on the interfaces. do you have any other ideas on how to monitor an ISP's circuit?

deverts · Answer

Only 1600?

To offer some more insight (at least from my perspective at a distance based on what I understand you are requesting), you are really looking for packet loss. But that means different things to different people. To some, discards can cause packet loss for certain, so that can be part of what you are looking for. For others, interfaces errors is a better indicator (my opinion). But then still for other people a complete end-to-end reflecting packet loss (any reason), hop-by-hop latency, and jitter provides the complete picture.

So, back to your original question...are you only looking for packet loss on the CloudGenix, or are you really needing a complete end-to-end visual of the customer experience? I think using UnDP to collect interface errors will provide you with the original answer, but end-to-end will require more tasks and data collection.

Another useful post:

CloudGENIX SD-WAN Solution Orion Monitoring and Alerting

rmullal · Answer

deverts​ thanks for your reply.

1. The only problem with using a netpath agent, is that we have ~1600 of these devices to monitor. I'm not sure how practical it is to have 1600 netpath agents. wouldn't i be restricted by resources on my polling engines?

2. so maybe i should try the ifInError/ifOutError OIDs?

3. We do have layer 3 Cisco devices on the other side of the CGX devices, but unfortunately do not have any licenses for VNQM.