SNMP requests to Cloud Service limitted/throtteled?

[Irrwitzer]

Hi guys,

I'm having some problems integrating some 240ish Meraki devices into our prometheus monitoring solution.

- SNMP Requests to snmp.meraki.com:$port with the given community string workED fine for ~a day. Then it seemingly stopped working. When I dug deeper I found, that a snmpwalk (with the suggested command line from the dashboard)  now takes a full 42minutes and 51seconds. For just ONE walk. Of course my monitoring requests every 120s now fail. Are there any limitations on how often we are allowed to query the dashboard SNMP service? What else could be the problem here?

- "local" requests to the individual devices work just fine, but I don't get information about PSU states, CPU usage, device temperature.... all of those would be essential for operation. Am I missing the correct MIB, or don't these devices (Switches and Access Points) expose those essential data points?

- The APs don't even report uptime? Seriously?

I wasn't a big fan of cloud managed devices from the start, but missing SNMP monitoring functionality with devices labeled with cisco is .... unexpected to say the least.

Any pointers in the right direction, or even to a fully working prometheus-exporter (utilizing the API or SNMP, I don't care) will be greatly appreciated.

Thanks in advance,

   Irrwitzer 

[alemabrahao]

SNMP Timeouts

In some circumstances it can take several seconds for our SNMP agent to initially query and build a database of devices for an organization. To give the agent enough time to respond, in the example commands above we have set a timeout of ten seconds. If you are using proprietary network monitoring software such as WhatsUp Gold, the SNMP timeout should also be configured, as the defaults are often too low.

https://documentation.meraki.com/General_Administration/Monitoring_and_Reporting/SNMP_Overview_and_C...

[Irrwitzer]

Thanks, I was aware of that section. But I don't see how 10s and 42minutes relate. My default timeout of 120s is also larger than 10s.... so what did I misunderstand?

[MartinS]

The SNMP implementation on Meraki devices is very limited - AFAIK all the monitoring development effort is going into APIs and webhooks. I found this which might be interesting, but it does look like Prometheus has extremely limited integration with Meraki's dashboard APIs - https://github.com/TheHolm/meraki-dashboard-promethus-exporter

[Irrwitzer]

Thanks Martin, I did find as well some days back. This piece of exporter code is rather old and the metrics exported are even less than what can be fetched via SNMP from the dashboard.

I might however use it as a starter for further extension if there's nothing more elaborate. I'm not a coder though, that's why I would prefer something existing to my own trial and error approach 😉

[MartinS]

Cool, there are other monitoring systems out there that have native Meraki API support - would be worth having a look at https://apps.meraki.io