Kepler dashboard is not found - 404

[msadofschi]

Hi,

While following the instructions to Deploy using Helm Chart in the "Post Install:" it is mentioned to "add the Kepler dashboard to Grafana", but the dashboard "404 - page not found".

Can the link be fixed or the dashboard added back?

Thanks!
Martin.

[sthaha]

@msadofschi thank you for reporting this. At this point, we are looking for contributors who are able to maintain helm chart. A sample dashboard that is usable with the new rewrite of kepler can be found here - https://github.com/sustainable-computing-io/kepler/blob/main/compose/dev/grafana/dashboards/dev/dashboard.json

Do you have the bandwidth to fix this?

[ExplorerRay]

I just tried this dashboard with current version of sustainable-computing-io/kepler-helm-chart.

But it seems not to be compatible with old Kepler image (release-v0.8.0 which is Chart appVersion).
And I also try to helm upgrade with v0.10.0-13-g8fd11d27 image which is newer. However, it reports kepler: error: unknown short flag '-v', try --help.

[ExplorerRay]

I use manifest/helm/kepler to install new version of Kepler, but still met a problem about dashboard.

I want to know how to get the uid of datasource in dashboard, the dashboard.json provides a static value, but in my environment, it seems to be a different uid which makes dashboard no data.

[Robbie558]

@ExplorerRay Importing the dashboard JSON in to Grafana does not work, and drops a large amount of the variable and panel configuration, most importantly the PromQL statements. I believe this occurs as Grafana fails to find the hardcoded datasource UID.

I did manage to manually fix this for a handful of panels, restoring the PromQL statements and legend configuration using the JSON. However I stopped as the use-case of the dashbaord appears more to compare kepler / kepler-dev / Node Exporter power reading, rather than report on power usage of nodes / pods / containers / processes.

Instead I used the archived kepler dashbaord JSON as my basis for a dashboard, correcting or entirely replacing the PromQL statements throughout.

[sthaha]

@ExplorerRay @Robbie558 Hopefully #2294 should solve the usecase. You are right the current dashboard in dev allows developers to compare the latest against their in-development version.

[ExplorerRay]

I just try the dashboard in my cluster (copy the JSON into grafana pod), everything looks well except the container part here. It still remains hard-coded datasource uid.

[Robbie558]

Thanks @sthaha. I have re-loaded the dashboard from the link provided by @ExplorerRay and am now getting results in all panels, which provides some of the information I was hoping for 👍🏻