SNMP/ICMP Support?

[haomingz]

Are you planning on supporting remote services, like SNMP or ICMP ?

[sparrc]

@haomingz at the moment there is not, telegraf is more meant for aggregating local data, but we could consider that for a future enhancement.

Could you provide more details on what you'd be looking for?

[pasancario]

I've some script working written in Python to retrieve information from network devices (mainly Traffic IN/OUT in interfaces).

It will be really usefull, if we can use Telegraf also as a daemon to work as a collector from remote devices publishing stats in SNMP.

Could it be possible create a Plugin, extending some SNMP library, and using Telegraph to multiple thread connection to each devices.?

[sparrc]

I don't see why not @pasancario, there is also the exec plugin, which needs to be documented better but basically allows for any JSON-emitting script to be collected by telegraf

[paulstuart]

I wrote a standalone snmp stats collector for influxdb 8.x, and am now updating it to work with 9.x (hopefully will be done next week).

https://github.com/paulstuart/influxsnmp

[JulienChampseix]

@paulstuart working like a charm ! @sparrc Hope to see this into telegraf as a plugin soon and that will be perfect and great improvement for telegraf tool ;)

[TheFlyingCorpse]

+1 on SNMP, would simplify greatly collecting from network / appliance devices.

Maybe some way to allow adding in new hosts to query via an API of sorts so it doesnt need to be one gigantic config file? 1000+ network devices with adds/removes every day.

[paulstuart]

@TheFlyingCorpse -- seems reasonable to me. The issue is how to do so cleanly - the hosts would have varying mibs to monitor, n'est pas? How do you envision this API looking?

p.s. I wrote this code for a project that has been sidelined, so I'm not even using these days myself :-(. I still want to make it useful for others so I'll do whatever I can as time allows.

[TheFlyingCorpse]

On the API, I'm not really sure. I'm used to Nagios/icinga with reload signal to make it reload its configuration, which is not really an API. It could easily be this in some form as long as it doesn't go and polll everything in the first 2 seconds then waits for say 298 seconds and does it again to everything, spread it out a bit.

What I'm struggling with in regards to SNMP is what is useful to gather and in what way it should be stored to fit with other similar data, what devices supply what information (EMC, pfSense, HP, BlackBox, Cisco++ and misc other appliances that support "SNMP"). There is even some vendors (Citrix that I know of) that have superspecial random OIDs generated on object creation in their appliances that can only be found in their GUI, SNMP walking from root does not show these.

[sparrc]

If any of you have comments on an SNMP plugin, please make them at #495, we should be able to get that merged soon and I want to make sure that it fits everyone's needs