Problems monitoring traffic on the ePMP force 300 client units

Over the last few weeks, we have been working on setting up a system to monitor and log the live traffic for out cambium ePMP customers.

Cambium has advised us to use cnmaestro for this task. We do have an installation of cnmaestro running, and it does monitor traffic under Performance->Throughput. Unfortunately, we have found that this graph works well only for small time ranges, but for larger ranges (multiple days), it averages the data to the point that it no longer gives any useful information.

We have tried to use other NMS systems (LibreNMS and PRTG) to monitor this data. The obvious way to do this would be to monitor an SNMP OID that shows the live interface traffic. However as we have found out, and confirmed with cambium support, the ePMP client units do not have such an OID, and do not use the industry standard OIDs.

The ePMP client units do have an OID that shows the total traffic that passed the interface since reboot, and cambium support advised us that this is what cnMaestro uses to monitor the device throughput. What i understand is that it continuously checks this value, and after it changes, it calculates the “live” traffic based on the total traffic change and time change.

I have replicated this by making a Python Custom Advanced sensor in PRTG. The only main issues that i ran into is that
(a), this is extremely inefficient as the NMS continuously has to run a script for every client unit that is being monitored. Not realistic to do this for over 2000 units that we have deployed without building out what would most likely need to be a server cluster.
(b) The total traffic OID seems to update at irregular intervals. Sometimes it takes less than 1 second and sometimes about 20 seconds. This means that to get acurate results, the script has to run for 30-40seconds to get a single data point. This dosen’t give enough granularity to troubleshoot certain network issues.

Does anyone have any other insight/experience/advice on setting up a system to monitor the ePMP client unit traffic?
Does cambium intend to implement a SNMP OID to allow simple monitoring of live interface traffic?

I would be interested in what others are doing other than paying out the yinyang for services like Powercode or maybe Sonar.

Back in the very early wisp days we pulled the data from SQL / Radius via Platypus accounting. But the numbers (customers and amount of data) got so huge that Platypus would just choke. We set up Argus (I think OpenArgus now?) and it was amazing what we could glean from it. Graphs, charts , lists on who, what , when, where and how much. But our network changed in a way (was broken up, segregated) that continuing with Argus wasn’t going to be practical.

Ultimately we ended up just not having the usage data. Other than pushing firmware updates I don’t really understand what it is CNMaestro is supposed to be good at… even if you pay for the non-crippled version I just don’t really, get it…

I have been thinking that going back logging usage with Radius again but I haven’t had time to really look into it and what little I have looked into it… it doesn’t look promising unless I want to pay a lot for a cloud solution or use open source solutions that haven’t been updated in years. I’m also not sure that the SQL choking up would be any better now then it was back in the day.

1 Like

We’ve been using CACTI to monitor. It’s not perfect, but it works pretty good. And As you say, the cnMaestro doesn’t monitor things in a useful way. Even when then sell the non-crippled features back to us… the non-crippled version still doesn’t monitor in a way to give me the useful info we need.

BTW, we also prefer an older version of CACTI from a couple years ago instead of the newer/fancier version - and it works OK for our diagnostic and graphing needs.

2 Likes

We use Cacti to graph a lot of stuff but like OP haven’t had good luck getting actual usage totals that are anything like actual usage totals. Really it seems the SNMP on ePMP is just as unreliable as the interface itself. As Rafal points out, it doesn’t update in a reliable usable way and we find that just randomly, once in a while it claims massive multi Gigabyte (byte not bit) spikes that just wrecks the averages. I’m thinking the only way to get anything useful is to pull it either from the Radius/SQL or the PPPoE server itself becaue ePMP radios are either brain damaged or pathalogical liers.

I’m setting up a machine now to give CNmaestro another go, just for updtating radios really and maybe GIS if they have the bugs worked out of that.

Thanks for the responses brubble1 and ninedd.
We will have a look at Cacti, as we were looking for a good NMS system (PRTG is quiet expensive for us)…
Unfortunately, we dont use PPPoE so thats not an option for us.

Very glad we are not the only ones that have seen that, we pole the 5-minute and 15-minute usages and they are a better way but still not perfect regarding the weird spikes.
Edit: we use Zabbix and SNMP monitoring through it all custom-built as there are not very many cambium templates for Zabbix surprisingly.

Preseem. One of the many useful features it provides us is client throughput. At $0.50 / user, it is well worth it for us.

We also use Preseem for this.