Telegraf memory sizing

Just wanted to get some specs for sizing a VM that is only running telegraf. I’ll be hammering on 1600+ nodes via SNMP for a few fields apiece roughly every minute or so, along with ping statistics to all the endpoints and exporting to InfluxDB.

I was thinking 2GB should do it for the VM (RHEL 7) as the process is currently only taking about 200MB of memory from what I see in top and I have half of the devices in this test instance already. I’m in a memory crunch so the less memory I can get this in at, the better.

I can’t directly answer your question, but memory should scale based on size of the metric_buffer_limit times the number of plugins enabled. If you have 2 snmp inputs enabled that could take ~ twice as much memory as a single one. If you want to see the worst case situation try making it so the output cannot send and Telegraf backs up.