Helm chart - install without grafana and prometheus about opencost HOT 11 CLOSED

yeah, we also already have prometheus and grafana in all of our clusters and it seems like a waste to have an extra instance for this.

from opencost.

@kivagant-ba, not today. It does have Prometheus as a hard dependancy. We are actively exploring other integrations though -- reach out ([email protected]) if you want chat in more detail!

As an alternative, you could install the bundled Prometheus and then export to your 3P monitoring system.

from opencost.

Hi @botzill, the main Kubecost application, which is a super set of this open source project, does indeed have the ability to ship without Prometheus & Grafana.

There are a couple nuances to implementing this, partially because of recording rules that Kubecust uses to remain performant on larger clusters. Right now, we generally recommend getting up and running with the bundled install first and then we can help you make the transition to your own Prometheus and Grafana.

Happy to help you do this -- reach out to [email protected] if you want to discuss further!

from opencost.

Can the app work without grafana and prometheus at all?

from opencost.

@kivagant-ba do you mind sharing a bit more?

The app can operate without Grafana today, although this limits some available views. It has a dependency on Prometheus though. Would you be looking to use a different database? We have discussed different integrations here so let me know what you're thinking!

from opencost.

I just don't have Prometheus since the company uses a third-party cloud monitoring system. That's why I'm curious if it's possible to use this very promising app without installing Prometheus.

from opencost.

Hi @dwbrown2. Can you share a little bit about Prometheus dependencies and how we can take those configs and put in our custom Prometheus installation? From what I see here https://github.com/kubecost/cost-analyzer-helm-chart/blob/master/cost-analyzer/templates/cost-analyzer-config-map-template.yaml#L8 we can easily change the endpoint to our custom installation of prometheus and alertmanager. Now, the next question is: do you have some custom cofigs when it comes to promethus itself, like did you add smth new here: https://github.com/kubecost/cost-analyzer-helm-chart/blob/master/cost-analyzer/charts/prometheus/values.yaml#L852 or is the default one? If there is added smth new I can take that and put input my promethus. For grafana, I see that there are these dashboards defined: https://github.com/kubecost/cost-analyzer-helm-chart/blob/master/cost-analyzer/charts/grafana/values.yaml#L178. Grafana allows to create a config map with dashboards and it will read it automatically. I think this is a good solution here as well, this way we can easily separate dashboards from grafana itself and use custom installations.

Hope this make sense :)

from opencost.

@botzill close :) you would want to modify https://github.com/kubecost/cost-analyzer-helm-chart/blob/master/cost-analyzer/values.yaml

I think it might be worth discussing before doing this in production. There are also recording rules and scrape configs that you need to pick up in order for this to perform properly on medium and large clusters.

Let's also plan to move continued conversation to https://github.com/kubecost/cost-analyzer-helm-chart since that's the relevant repo.

from opencost.

New issue in Helm chart created if there's interest in further discussion!

kubecost/cost-analyzer-helm-chart#7

from opencost.

We already have prometheus and prometheus-node-exporter running in our cluster. Install kubecost brings prometheus-node-exporter as well which is always in Pending state, because its port has been occupied. Can I install kubecost without prometheus-node-exporter ?

from opencost.

You definitely can! Check out the kubecost helm chart https://github.com/kubecost/cost-analyzer-helm-chart/blob/develop/cost-analyzer/values.yaml#L583

from opencost.