The flush operation took higher than the expected set push interval

Grafana k6
3

Hi @MattK6,
the requests for flushing metrics to InfluxDB are taking longer (1.6s in your reported case) of the set flush interval (default to 1s) so it means that k6 is probably flushing more metrics than InfluxDB can really handle.
Do you see a high value of CPU and/or RAM for your InfluxDB instance? Are you using the same machine for running k6 and InfluxDB or are they separated?

I list here some steps to try for getting a better response time from InfluxDBv1:

  • You could set only the SystemTags that you really need for your analysis. This can help for reducing the number of metrics that k6 is generating and InfluxDB have to ingest.
  • Tune the options of the k6 output for InfluxDB:
    • You could consider excluding some tags to be indexed K6_INFLUXDB_TAGS_AS_FIELDS that should speed up the ingestion phase for InfluxDB. Note that not indexed tags could be slow when you need to query them.
    • Reduce the K6_INFLUXDB_PUSH_INTERVAL and increase the K6_INFLUXDB_CONCURRENT_WRITES options for flushing batches with a smaller number of metrics. You probably will need to check the logs with the –verbose option for setting the right values.
  • Try to use Telegraf for aggregating some metrics before sending them to InfluxDB.

If none of the previous suggestions works for you then you could try the new InfluxDBv2 extension and see if it’s able to ingest your metrics faster.

1 Like