Best way to utilize Grafana Cloud with an already preexisting architecture on AWS EKS

We currently self-host our LGTM stack and has been working great for us; however, we have found it hard to continuously maintain it with our small team. We currently host and configured all from AWS EKS but would like to know how developers utilize Grafana Cloud to help ease the burden.

I have heard they use Grafana Cloud for Visualization while LTM are maintained locally on their own K8s cluster. Is this very common to keep costs moderately low? Would like your feedback.

That’s opinionated question “the best way”. What’s the best for another company may not be best for you. I see your problem is management of stack, so you think SaaS (Grafana Cloud) can be better (but you can be price sensitive, so again it may not be best for you).

So first define what’s is that “the best” in your case, define budget and model differernt approaches.

I work with both approaches:

  • Grafana Cloud: I’m happy that I don’t need to lift a finger, but I have to pin my versions, otherwise I may be firing tickets for support, because they have unattended upgrades (even their the slowest release channel is too fast for my case)
  • on-prem: many local resources, which can’t be exposed on the internet (which was a showstopper before recent Grafana PDC feature)

So who knows what’s the best for your case? Only you.

Usually people who use Grafana Cloud for visualization will also use the data storage options offered by Grafana Cloud.

If you’re only just starting out or migrating all at once seems like a daunting task, you can take it one step at a time & choose the parts which work for you.

For example: if your log volume is starting to slow you down, you can replace your locally hosted Loki instance with Grafana Cloud Logs while keeping everything else in your setup exactly the same. Afterward you can continue with Metrics, Traces etc. or keep your existing setup until you run into scaling issues there.

Because Grafana can pull in your data from everywhere, your end users won’t notice the difference.