Issue with InfluxDB performance: cache-max-memory-size exceeded

Grafana k6 OSS Support
1

Hi,

We are facing this issue on InfluxDB when we run more of 5000 request per second.
level=error msg=“InfluxDB: Couldn’t write stats” error=“{"error":"engine: cache-max-memory-size exceeded: (1075899908/1073741824)"}\n”

This is my configuration :
k6 v0.30.0
(2021-01-20T13:14:50+0000/2193de0, go1.15.6, linux/amd64)
Server : Centos 7.0

DB :
InfluxDB
1.7.9
RAM : 64GB

There is some best practice to avoid this issue? some configuration to modify in InfluxDB?

Thanks in advance

2

Hi, welcome to the forum :slight_smile:

This question is better suited for an InfluxDB forum than here, but I’ll give it a shot.

You can try increasing the cache-max-memory-size value from the default of 1G to whatever makes sense for your use case.

If you’re starting it with Docker Compose, you can specify it with the INFLUXDB_DATA_CACHE_MAX_MEMORY_SIZE environment variable, e.g.:

services:
  influxdb:
    image: influxdb:1.8
    networks:
      - k6
      - grafana
    ports:
      - "8086:8086"
    environment:
      - INFLUXDB_DB=k6
      - INFLUXDB_DATA_CACHE_MAX_MEMORY_SIZE=4g

Also see this InfluxDB forum answer for other tuning ideas, including making use of fast storage.

In general tuning InfluxDB for better performance at large scale is a difficult problem, since k6 generates so much metric data.