How many concurrent users/sessions can Grafana handle?

We are looking into using a combination of Grafana and Influx for monitoring. In our setup we would have 10.000’s of users logging in to view dashboards. Some of them will also be doing some custom queries but the majority will just look at some predefined dashboards.

Has anyone had any real experience with 10.000’s of concurrent sessions?
From my research I understand that Grafana does not cache any results so that would mean that each individual session will make their own request to the database at every refresh interval. This would then mean that the Influx database would get bombarded with 10.000s of requests.

I am unsure if this is doable.
Any feedback would be greatly appreciated.

It depends on your budget and infrastructure. You can scale horizontally both products easily: Grafana and InfluxDB (InfluxDB enterprise version only). So when you hit the limit, then you will add new node/container/VM/…

Thank you for the response Jan.

We are correct then in assuming there is no way to only have Grafana query the database once and then distribute the results to all sessions that currently have that dashboard open?

Do you have any idea of how many instances of Grafana/Influx would be needed to be able to serve 10.000 concurrent sessions?

You are asking for push - yes that’s possible but there is only one datasource, which supports it - snap app. Others are using classic pull approach. Nobody is stopping you to build own cache layer (Varnish) in front of InfluxDB.

Sizing: there is many input variables - network, cpu, mem, query duration, LB, … so your definition of 10k concurrent session is very vague for proper sizing. Why you don’t test it with your real data/infrastructure?