Grafana/Influxdb live monitoring : With distributed load testing facing duplicate data points sends to Influxdb

Gatling (Open-Source)
1

Hello Stephane,

We have configured influxdb with Grafana for Live Monitoring and it works good for load injection from single machine. Numbers shown in Grafana dashboard is matched with gatling html report.

But when we do inject load from multiple machines there we are noticing issue. The numbers are not at all matched.

Ex: From 1 machine we get a throughput of 50k per second.
With 2 machines gatling report shows 100k, but influxdb/grafana is showing around 48k.

Influxdb & grafana is hosted in single machine ec2 instance R4 (2 cpus,16 gb memory), where TCP tunings are done as recommended in Gatling website. There is no peak in memory and cpu from influx db side.

I did a research around this issue and came to know about duplicate data points handled in influxdb

https://docs.influxdata.com/influxdb/v2.0/write-data/best-practices/duplicate-points/

I suspect when we run the test from multiple machines there could be chances of sending same data points from gatling to influxdb. If duplicate is an issue by adding tags in measurement (influxdb) we can handle this scenario. I don’t see anyway to pass tags like ‘Region -1, Region-2, Machine 1,Machine 2’ from gatling side.

Could you please suggest/share your thoughts on this behaviour ?

Thanks,
Ram

2

Well, distributed load test is featured in FrontLine, not in Gatling OSS.

3

Yes distributed load test is featured in FrontLine. Sorry i used the wrong term, we are scaling out in OSS from multiple machine.

4

Generating load from multiple injectors is exactly what “distributed” means :slight_smile:

Gatling OSS is meant for single node usage.
Things on the OSS side not working properly for multiple nodes is entirely possible as it’s out of scope.
Multiple node usage is covered in FrontLine.