SNMP Monitoring with RFC5907 MIB

[avluis]

Would like to see monitoring via SNMP (v2/v3 preferred) as well as minimal ability to configure what it presents (optionally support changing community settings)

[dave4445]

Are you looking for a specific NTP MiB or just genetic system ones?

[avluis]

If rfc5907 {NTPv4} is considered alongside standard (system) OIDs I'd be a happy camper

[dave4445]

NOTE: multiple people are now asking for this

[lump]

I just ordered an NTP250 for my homelab. I have been perusing the documentation in preparation, and so far, I like everything about this appliance except that it seems to be missing SNMP and a legit serial port (which I believe might be a better place for a PPS signal than an unprotected circuit). SNMP would be a great addition that I would most definitely use to graph from Observium and alert thresholds from Nagios.

[dave4445]

Thanks Troy, I started on this a few months ago and adding system MIBs isn't a big deal, however finding something for rfc5907 was a bit more troublesome. Threr's a PoC AgentX that implements parts of it out there from a GSOC project but it's not really useful as is. Implementing this MIB from scratch would be quite time consuming

[lump]

Wow, I was just chiming in with my +1 for the record. Now I look forward to this product even more because talking with the developer at this level makes it even more amazing. Thank you for your responsiveness!

I agree, to write it from scratch would be painful. I was hoping you were running ISC NTPd under the covers and could simply turn on the ntpsnmpd daemon after installing net-snmp and turning on the master agentx socket. But if you're doing closed-source stuff, I'm not sure what to say.

[dave4445]

Yes, but if you look closely at that ntpsnmpd it's just a proof of concept and very incomplete, only implementing a few of the MIB objects. hasn't been touched in 8+ years.

[lump]

Oh, sigh, that's sad. You're right, it is basically an abandoned Google Summer of Code project.

Now that I think of it, my Observium metrics graphing uses a fork of the check_mk agent instead of getting NTP stats over SNMP.

Not that I don't like the NTP250's graphs. I just like to centralize metrics graphing.

But, to get these kinds of stats from an appliance, it might be better to use a standard like SNMP instead of having to support any number of kinds of agents. If I am allowed to keep asking, the next thing I'd ask for is a page that lists stats that Prometheus can scrape (and then that Grafana can easily graph). But I can probably write a transform for the already provided JSONs for Prometheus. SNMP is a lot lower level.

[dave4445]

SNMP is now available in v2.7 minus the NTP mib

[lump]

Wow, thank you! It looks like my device already got the update!

I enabled SNMP, and the device is answering! I already added it to Observium. It is nice to at least have the system resource graphs for now, thank you.

[avluis]

Received the update recently and enabled the feature; thank you for implementing this!