I’m looking at the InfluxDB line protocol output generated by Icinga, e.g. for an icmp measurement, and it doesn’t look right to me.
As per InfluxDB definition, it should look like this, with metrics recorded in the field set.
icmp,host=my-host rta=0.177983,pl=0 1741247075000000000
Instead, it looks like this, with all fields set to value and metric name referenced within the tag set instead.
icmp,host=my-host,metric=rta value=0.177983 1741247075000000000
icmp,host=my-host,metric=pl value=0 1741247075000000000
The resulting data is more difficult to query (need to filter on “metric” tag instead of simply referring to the field) and inefficient to write (generates more traffic).
Those examples are of course simplified to better illustrate my point, there would be more tags in real life. I cannot find anything about fields in the exporter documentation, only how to set measurement and tags. My workaround is to use a Telegraf InfluxDB listener as a “proxy” to pivot/merge this data into the proper format before writing it to InfluxDB.
I was wondering if this is a deliberate design decision (e.g. to include units in the output)?