8 billion values, influxd uses 50 Gb ram - is this normal?


I have loaded influxdb with roughly 13000 series which contain about 8 billion data points.
Currently, influxd uses quite a lot of RAM, on server that has 7 Gb, RSS size is 6 Gb (so about
all RAM is being used by influxd). Virtual process size is 50 Gb.

Influxdb version is 1.3.3. I attempted to reduce memory usage, so I have set index-version = “tsi1” in influxd.conf and
changed shard duration from 1 week to 1 year. But there were no dramatic changes in ram usage.

Is this normal ? Should I just throw more hardware at this ?

I cannot downsample the data, it is all needed for analysis in it’s original resolution.

Yes, this is typical memory usage for Influx. You must have realized that it will temporarily grow with the queries. If procuring hardware/software isn’t an issue you can also look at the Clustered Enterprise solution for a distributed storage/query solution.