Skip to content
This repository has been archived by the owner on Mar 19, 2021. It is now read-only.

grafana dashboards K8s Node doesn't display ? metrics #56

Open
shiyiwenv opened this issue Jan 23, 2019 · 8 comments
Open

grafana dashboards K8s Node doesn't display ? metrics #56

shiyiwenv opened this issue Jan 23, 2019 · 8 comments

Comments

@shiyiwenv
Copy link

image

Everything else is normal. What's the reason?

@tangjiaxing669
Copy link

me too.

@sosoriov
Copy link

Same here!
I checked the metrics and panel is link to node_cpu metric which is not available in my Prometheus,

2019-01-24 10_02_05-grafana kubernetes app

@wguto
Copy link

wguto commented Jan 24, 2019

ye it is an old metric, check my PR, it is fixed there. The list I couldnt figure out why it is not showing =[

@sensay-nelson
Copy link

sensay-nelson commented Jan 31, 2019

@wguto oddly, using your PR I have the list working, but none of the other metrics :( . I can swear that this data was working the other day though, and then stopped working as I fixed the Container metrics.

 node-exporter:
    Image:         quay.io/prometheus/node-exporter:v0.16.0

Kubernetes:
Server Version: version.Info{Major:"1", Minor:"8", GitVersion:"v1.8.7", GitCommit:"b30876a5539f09684ff9fde266fda10b37738c9c", GitTreeState:"clean", BuildDate:"2018-01-16T21:52:38Z", GoVersion:"go1.8.3", Compiler:"gc", Platform:"linux/amd64"}

Kubernetes plugin version 1.0.1

screen shot 2019-01-31 at 10 16 29 am

edit: I found the source of my issue was that somewhere along the line I had deleted and re-imported the dashboards, but they picked up the old specs rather than the new ones. I'm chalking it up to Grafana's odd way of handling dashboard versions.

@sosoriov
Copy link

I would say it depends on your Prometheus configuration. Definitely the metrics had change and now you must use the new names, however, depends on your configuration you will get different labels in your metrics, for instance. I saw in @wguto PR that you filter based on instance label but in my case I have kubernetes_node label. So I would suggest @sensay-nelson that double check the values that comes from Prometheus and compare them with your Grafana template.

Some examples of the new metrics names here

https://prometheus.io/docs/guides/node-exporter/

Before it was node_loadxxx now the metric is node_loadxxx_total check the Prometheus guidelines

@sensay-nelson
Copy link

it's all a bit mysterious, but I agree that the metric name and the node label appear to shift a bit depending on setups. Using node-exporter v0.16.0 the metric names from Wguto's request were correct (node_cpu_seconds_total), but the '$instance' labels are '$node' labels in my setup.

@shiyiwenv
Copy link
Author

it's all a bit mysterious, but I agree that the metric name and the node label appear to shift a bit depending on setups. Using node-exporter v0.16.0 the metric names from Wguto's request were correct (node_cpu_seconds_total), but the '$instance' labels are '$node' labels in my setup.

How to setup '$instance' labels are '$node' labels ? my image update to node-exporter v0.16.0 ,That's it (node_cpu_seconds_total) 。

@wguto
Copy link

wguto commented Feb 13, 2019

@sensay-nelson @sosoriov
Another thing that I notice when working with other dashboards for more in depth network stuff, like the dashboard 1860 on the grafana repo, it is better to use the node-exporter V0.15.2. For some reason it has LOTS of more metrics for heavy network stuff like tcp stuff.
Now I switched back to v0.15.2 and this dashboard is still working, of course I have to change the metrics all again to the old format.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

5 participants