Did some testing…
- 2 Telegraf containers collecting data from routers and sending to Kafka output.
- Each telegraf.conf file is configured to collect data from 5-6 routers.
- One InfluDB container that consumes data from the Kafka topic that Telegraf is producing to.
I enabled the [[inputs.internal]] plugin on one container and now see the following measurements:
> show measurements
Ran the following query to calculate the rate of metrics written per minute:
> select non_negative_difference(last("metrics_written")) from "internal_write" where time >= now() - 10m group by time(1m)
I believe this gives me the metrics written per minute. The huge spike in the rate I believe is caused by how many metrics Telegarf counted at the moment it was writing to the output, which is Kafka in my case. Not really sure how to smooth those numbers out.
I then enabled the [[inputs.internal]] plugin on my second Telegraf container, which is also writing to the same Kafka output, but I did not see the rate increase dramatically from what it was when I only enabled the plugin on the first container.
I hope I’m testing this correctly. If someone at Influx can confirm and/or correct this testing that would be awesome.