Influx V1 offered metrics interface via the
input.influxdb telegraf plugin which resulted into a nicely structured schema.
Influx V2 offers a metrics interface
:9999/metrics in prometheus format. I use now the
input.prometheus telegraf plugin to ingest this into influx again. Telegraf generates about 1600 metrics per interval, all with same time stamp, with varying key sets and always a single field. And InfluxDB unions this into a single point with a large key set and a large field set. Simply what one gets converting prometheus format data.
After a few hours run time on a low load example system the schema looks
SHOW SERIES CARDINALITY FROM influxdbv2 --> 1375 SHOW TAG KEY CARDINALITY FROM influxdbv2 --> 29 SHOW FIELD KEY CARDINALITY FROM influxdbv2 --> 161
Doesn’t look like a schema InfluxDB is designed for.
So I really wonder what the rationale was to use the prometheus format for